optimal treatment policies for pelvic organ prolapse in women · 2019/9/4 · author: optimal...
TRANSCRIPT
Submitted tomanuscript
Optimal Treatment Policies for Pelvic OrganProlapse in Women
Yueran Zhuo, Senay SolakIsenberg School of Management, University of Massachusetts Amherst
[email protected], [email protected]
Oz HarmanliDepartment of Obstetrics, Gynecology, and Reproductive Sciences, Yale School of Medicine
Keisha A. JonesDepartment of Obstetrics and Gynecology, Tufts University School of Medicine Baystate Medical Center
Pelvic organ prolapse (POP) is a very common gynecological disorder greatly affecting the quality of life
(QoL) of females in the society. Studies have shown that approximately 40% of women in the U.S. have POP
that requires medical intervention. In clinical practice there are four treatment options for POP: watchful
waiting, conservative treatment, reconstructive surgery, and obliterative surgery. In this study, we utilize
practical data obtained through surveys and clinical literature, and develop a model to help physicians
and POP patients dynamically select treatment options in order to maximize a patient’s expected future
quality of life. Results from the model are presented in the form of optimal policy tables, which can be
used by physicians and patients by entering several attributes as inputs, such as the patient’s age, current
observed QoL defined by the severity of POP symptoms, preferences over preserving coital function, and
other potential restrictive conditions. The results show significant socio-economic incentives for potential
utilization of optimal treatment policies in POP treatment. We estimate based on QoL to dollar conversions
that the expected value of improved QoL for an individual POP patient is around $10 thousand. When
aggregated over the entire POP population seeking treatment, the total annual expected value for the society
is at least $675 million under medium level valuations of QoL. Moreover, the optimization based policies
imply an overall reduction of about $200 million in annual POP treatment costs, which corresponds to a
savings of about 5% in these costs.
Key words : healthcare, treatment selection, pelvic organ prolapse, quality of life, Markov decision process,
dynamic programming
1
Author: Optimal Treatment Policies for Pelvic Organ ProlapseArticle submitted to ; manuscript no. 1
1. Introduction
Pelvic organ prolapse (POP) is a common gynecological disorder impairing many women’s quality
of life (QoL). Prolapse results from weakening of the muscles and ligaments supporting the pelvic
organs. A POP patient suffers from symptoms such as constant pelvic pressures, low back pains,
bleeding, incontinence and constipation. As a chronic disease, POP affects a large population of
females in the world, especially senior women with childbirth experience. Studies have shown that
up to 76% of women in the U.S. have shown some levels of POP symptoms, and up to 19% require
some kind of medical intervention (Barber 2016). Overall, approximately 200,000 inpatient surgical
procedures are performed every year in the U.S. to treat POP, incurring costs of over $1.5 billion
(Jones et al. 2010). Moreover, with the increasing life expectancy worldwide, the impact of POP
on quality of life has become even a bigger concern, as the condition is more common in elder
patients. The 2013 World Population Aging Report, released by the Department of Economic and
Social Affairs of United Nations, indicates that the number of people above age 60 will increase
rapidly in the next four decades to reach 21% of the entire population by the year 2050. This is
expected to result in a 45% increase in the demand for POP treatment (Jones and Harmanli 2010).
Given these observations, it is important to consider the various options available to treat POP
patients and identify practical guidelines for selecting treatment options for such patients, as there
is no such quantitative-based guidance available in the literature. In this paper, we study this
problem using a stochastic dynamic model that aims to maximize the expected future quality of
life in patients with POP, where the results are aimed to provide insights to physicians in making
clinical decisions when treating patients with POP.
1.1. Problem Description and Research Questions
A number of options has been developed and implemented in clinical practice when treating POP.
These options can be classified into four categories: watchful waiting, conservative treatments,
reconstructive surgery, and obliterative surgery. Watchful waiting involves no specific treatment
action other than observing the progress of POP symptoms. Conservative treatments include non-
surgical options for POP, such as the use of a pessary to support the pelvic organ, pelvic floor muscle
exercises, weight-loss, and other treatment that aims to relieve POP related discomfort. Recon-
structive and obliterative surgeries represent the surgical options for POP, but the two approaches
and subsequent consequences differ. Reconstructive surgeries restore pelvic organ’s anatomy and
provide resuspension of vaginal vault. Reconstructive procedures typically have longer operating
times and pose a higher risk of complications, but postoperative coital function is preserved. Oblit-
erative surgeries, on the other hand, permanently close the vagina. In comparison to reconstructive
Author: Optimal Treatment Policies for Pelvic Organ Prolapse2 Article submitted to ; manuscript no.
surgeries, obliterative procedures require shorter operating times and have a decreased risk of com-
plications. Overall, obliterative surgeries have success rates near 100% (Fitzgerald et al. 2006),
which is significantly higher than that of reconstructive surgeries, as it is estimated that 27% of
reconstructive surgeries might require follow-on operations (Harmanli 2009).
When discussing POP treatment options with patients, gynecologists consider factors such as the
patient’s age, expectations, health state, degree of bother, and desire to maintain coital function.
Based on such factors, a patient with lower POP stages may prefer more conservative treatment
options before considering a surgical intervention. For patients with higher stages of POP, as the
degree of discomfort is higher, a patient may be more amenable to some surgical intervention. On
the other hand, there are no general guidelines that a physician can use to recommend a specific
treatment option to a patient. Physicians make recommendations based on patient preferences and
clinical experience without referring to a specific quantitative analysis based guideline.
Our objectives in this paper are to identify optimization-based treatment policies to serve as
a reference for physicians, and also to assess the value of such optimization-based policies by
comparing these with the current policies used in practice. More specifically, we utilize both clinical
data and survey data, and try to answer the following research questions: (1) Given the available
data on treatment options and their effectiveness over time, what are optimization based treatment
policies that maximize a POP patient’s expected future quality of life? (2) How do the treatment
decisions in these policies differ from those currently used by physicians? (3) What value can
potentially be generated through the use of optimal policies as decisions aids by physicians? (4)
What is the cost impact of optimization based treatment policies when compared with the current
practice?
Our approach to answer these questions involves a Markov decision process (MDP) model that
captures the dynamics of the decision process in POP treatment by utilizing physician surveys and
available clinical data. The model is used to derive optimal policies under different settings, and
these policies are then compared with the current clinical policies as identified through survey data
collected from physicians.
1.2. Review of Relevant Literature
MDP-based approaches have been extensively used in studying a variety of medical decision-making
problems, especially for certain major life-threatening illnesses (Schaefer et al. 2004, Alagoz et al.
2010). Some of these applications include organ transfers (Ahn and Hornberger 1996, Alagoz et al.
2004, 2007a,b), cancer screening (Maillart et al. 2008, Chhatwal et al. 2010), treatment of diabetes
(Denton et al. 2009, Kurt et al. 2011, Mason et al. 2012, Kirkizlar et al. 2013), human immunode-
ficiency virus (HIV) infection (Shechter et al. 2008), and cardiovascular diseases (Hauskrecht and
Fraser 2000, Montgomery et al. 2003).
Author: Optimal Treatment Policies for Pelvic Organ ProlapseArticle submitted to ; manuscript no. 3
Of these papers, Alagoz et al. (2004) study the timing decisions for liver transplans under insuf-
ficient cadaveric organ supply, where the authors build an MDP model to maximize the expected
quality of life for patients. Similarly, Shechter et al. (2008) discuss the time to initiate HIV treat-
ments for patients in critical status. Another study of this type is by Denton et al. (2009), where
the optimal timing for statin therapy is determined to treat Type-II diabetes in order to minimize
cardiovascular risk according to a patient’s age, gender, and metabolic state. Similar to our paper,
all these studies utilize the concept of QoL as the measure to be optimized. On the other hand,
the medical decisions considered in the studies above are one-time treatment decisions rather than
consisting of multi-period treatment plans, as in the case of POP treatment which is prolonged
over the lifetime of a patient. One study that assumes a multi-period treatment plan is by Chhat-
wal et al. (2010), where a patient’s risk of breast cancer is evaluated in regular intervals, and the
decision of whether or not to perform a biopsy is made based on the patient’s observed condition.
Using a finite-horizon MDP model, the authors identify an optimal policy based on the patient’s
age and breast cancer risk level. Similarly, in our problem framework, the treatment decisions are
also made based on an annual examination of the POP patient, but the available actions involve
four different treatment options. The multi-period treatment plan and the availability of multiple
actions at each decision epoch add complexity to both the modeling and analysis in our problem
framework.
As a non-fatal disease, POP has been rarely studied in the field of operations management, and
moreover, MDP-based approaches have not been previously utilized in the gynecology area, despite
its successful applications in other medical fields. However, one cannot neglect the significance of
the disease and its growing social implications especially due to an aging population. To the best
of our knowledge, our work is the first study to address treatment planning for POP using an
optimization approach that provides patient-specific long-term treatment recommendations. Thus,
as a novel and data-based application of operation management methods in healthcare, our study
complements the existing MDP-based models on impactful disease treatment management.
With respect to the existing literature in the gynecology field on POP treatment, most studies
focus on the characteristics of POP and associated treatment options. Samuelsson et al. (1999),
Swift et al. (2005), and Jones et al. (2010) provide some general statistics about the population of
POP patients, while Barber et al. (2001) introduce two measures for POP’s impact on quality of
life. Ellerkmann et al. (2001), Nygaard et al. (2004), Handa et al. (2004), and Gutman et al. (2008)
study the natural progress of POP development, and evaluate how such progression impacts a
patient’s quality of life. These studies help characterize the model assumptions and input structure
in our research. Wheeler et al. (2005), Barber et al. (2006), Barber et al. (2007), Nguyen and
Burchette (2008), and Harmanli (2009) each study a specific POP treatment option, and report the
Author: Optimal Treatment Policies for Pelvic Organ Prolapse4 Article submitted to ; manuscript no.
effectiveness of the corresponding treatment. In addition, another study relevant for our work is by
Jelovsek et al. (2015). In that study, the authors describe the characteristics and application scopes
for each of the POP treatment options, and then provide specific qualitative suggestions for POP
patients as to which surgical procedure to use at a given age for different levels of POP symptoms.
With some degree of consistency with our work, the conclusion of that paper can be seen as a
qualitative counterpart of our quantitative study, which can help physicians better interpret our
policy findings.
1.3. Contributions and Major Findings
Our study contributes both to the current state of research and clinical practice for POP by
providing insights for treatment. As part of the major findings of our research, we draw the following
general conclusions for optimal POP treatment policies: (1) For younger patients up to age 50,
watchful waiting is always recommended when the patient’s QoL level is above 7 based on a scale1
of 1 to 10 for QoL, with 10 being the healthiest case. This QoL threshold gradually decreases at a
rate of around 1 level per 16 years as the patient gets older; (2) Surgical options are recommended
for patients with QoL levels below 6 for all ages; (3) The age threshold to prefer obliterative surgery
over reconstructive surgery is 74; (4) Obliterative surgery should not be considered when the desire
to preserve coital function corresponds to more than 40% of a patient’s quality of life.
Comparing these policies with the current practice, we quantitatively conclude that: (1) The
proposed policies are expected to provide QoL improvements of between 9% and 18% over currently
utilized physician policies; (2)The expected annual socio-economic value of optimization based
policies can be at least $675 million when the treatment seeking POP population in the U.S. is
considered; (3) Optimal policies imply an estimated savings of around $200 million in annual POP
treatment costs while at the same time improving the QoL of patients.
The remainder of this paper is organized as follows. We develop the MDP model for selecting
treatment options for POP in Section 2. In Section 3.1, the MDP model is used to identify optimal
treatment policies based on available clinical data. In Section 3.2, we compare the optimal treatment
policies with the current physician policies, and quantify the potential value generated by our
proposed policies. In Section 4, we summarize the policy implications of our findings, and in Section
5 we conclude the paper by some final remarks.
1 We describe in Section 2.1 how this scale is developed and how the current QoL of a patient is measured based onthis scale.
Author: Optimal Treatment Policies for Pelvic Organ ProlapseArticle submitted to ; manuscript no. 5
2. Methods
In this section we modeling the Decision Process for POP Treatment Selection. Consider a patient
seeking treatment for POP symptoms. The decision process for treatment selection starts with the
evaluation of relevant patient data, which include the patient’s age and other measures defining
her quality of life, mainly as it relates to POP-related discomfort. After consideration of all rele-
vant inputs, a treatment option is recommended by the physician. Once the treatment option is
implemented, the effect of the treatment is then observed and the post-treatment quality of life is
evaluated. If necessary, further treatment options can be considered based on this evaluation. Such
decision cycle continues until the termination of a planning horizon, which usually happens when
the patient’s status remains stable or when an age limit is reached. In the following subsections we
describe in detail how the components of an MDP model can be developed in the POP treatment
context based on the decision process described above.
2.1. Decision Framework
POP is mostly common in middle to older aged women, especially among those with childbirth
experience. Most empirical studies on POP consider subject groups of ages between 45 to 85, as rare
cases of pelvic organ prolapse at younger ages are seldom symptomatic (Samuelsson et al. 1999).
Hence, we choose an age range of 45 to 85 years as the planning period for selecting treatment
options for POP. We further assume that a gynecological exam is performed annually to assess
a patient’s condition. Based on the assessment at each exam, a treatment option is selected and
implemented. This dynamic treatment decision process is illustrated in Figure 1, where an initial
treatment is followed by a period of observation of the resulting disease progression before another
treatment plan is decided at the next decision epoch. The process goes on repetitively until the
end of the treatment planning horizon.
2.1.1. Modeling Inputs for the Decision Process Based on interviews with physicians,
it can be noted that the treatment decisions for POP are mainly based on the patient’s age and
three other high level measures: the current quality of life of the patient, the surgical history of the
patient, and the patient’s preference over preserving coital function. We describe these measures
in detail in the following paragraphs.
Modeling Quality of Life. When appropriately devised, a measure of patient’s current quality
of life captures the severity of POP symptoms, as well as the impact of the resulting distress on
daily life. While quantification of quality of life in this context is somewhat of a challenge, the
gynecology literature contains some symptom-based quality of life measurements. Two of these
metrics have been widely adopted by physicians, namely the scores from the Pelvic Floor Distress
Author: Optimal Treatment Policies for Pelvic Organ Prolapse6 Article submitted to ; manuscript no.
Inventory (PFDI) and the Pelvic Floor Impact Questionnaire (PFIQ), which are surveys completed
by POP patients before and after treatment to assess POP-related distress symptoms. These two
metrics evaluate the quality of life of a POP patient by assessing the degree of discomfort and other
problems caused by pelvic floor symptoms. Three major concerns due to POP are considered in
both measures: impacts on daily activities due to pressure in pelvic area, urinary incontinence, and
fecal incontinence. The PFDI score is a summation of the sub-scores for three sub-questionnaires
corresponding to the three areas of concern above, where each area is scored using a range of 0 to
100. Similarly, PFIQ also has three sub-scales on these concerns, each with a scale between 0 and
100. The summation of these three sub-scales presents a total PFDI or PFIQ score between 0 and
300 with higher scores being indicative of more severe levels of symptoms.
Barber et al. (2001) compare PFDI and PFIQ scores, and conclude that these two scoring systems
have significant internal consistency with a high degree of correlation. Hence, in our study we
use existing empirical studies on POP that utilize one or both of these measures. We do this by
converting the PFDI and PFIQ scores into a new scale ranging from 1 to 10, based on a rescaling
of the original PFDI and PFIQ scores reported in the POP related empirical studies.
Modeling Surgical History. The second important criterion in deciding on a treatment option
is the surgical history of the patient. This is relevant because some patients may end up going
through multiple reconstructive surgeries as part of their treatment process due to recurrence of
symptoms after surgery. On the other hand, due to the theoretical risk of complications associated
with multiple surgeries, the total number of reconstructive surgeries a patient can have is limited.
Olsen et al. (1997) report that only in extreme cases a patient should be going through more
than three reconstructive surgeries. Similarly, for a patient with a previous obliterative surgery, no
further surgical options are feasible due to the permanent closure of the vaginal opening. Hence,
the number of obliterative surgeries a patient can have during the treatment period cannot exceed
one. Based on these, we keep track of the number of surgeries for a patient in our model, and limit
the number to three for reconstructive surgery, and to one for obliterative surgery.
We also note that the surgical history is not the only factor constraining surgery choices of a
POP patient. The patient’s health condition, preparedness and personal preference on the available
types of treatment also play a role in choosing a surgical procedure. Therefore, the categories
used to distinguish surgical history can also be used to characterize these special conditions of a
patient. For example, a patient who is determined to preserve coital function would be considered
equivalent to a patient who had gone through obliterative surgery, as in both cases the obliterative
surgery option is not feasible. In Section 3.1 we perform analyses based on such considerations.
Author: Optimal Treatment Policies for Pelvic Organ ProlapseArticle submitted to ; manuscript no. 7
Modeling Patient’s Preference of Preserving Coital Function One particularly important issue
about treatment of POP is the impact of treatment options on the lifestyle of a patient. According
to Ratner et al. (2011), 30% of women between the ages of 55 and 75 remain sexually active.
This ratio is around 17% for women between the ages of 75 and 85 (Matthias et al. 1997). While
conservative treatments and reconstructive surgeries result in some temporary inconvenience in a
patient’s daily life, they do not have long-term impacts on coital function. On the other hand,
obliterative surgery closes the vaginal opening and inhibits any coital activity permanently. Despite
the prevalence of such concerns in treatment selection for POP, the issue has been seldom discussed
in the gynecology literature. Jelovsek et al. (2015) suggest that obliterative surgery should be
conducted on a patient only when the patient cannot tolerate more extensive surgeries, and/or
does not plan any further coital activity. However, the definitions of extensive surgery tolerance
as well as inclination for coital activity are somewhat vague, and typically are not considered in a
systematic way in practice. It is reported in a post-surgery survey that 9% of POP patients have
regrets after obliterative surgery (Wheeler et al. 2005).
In this study, we introduce a preference index Pn(t) to capture the desire of a t−year old POP
patient to preserve coital function. We then perform sensitivity analysis around this preference
index to determine the impact of a patient’s choice on this concern. We assume that before deciding
on a treatment option, a patient will be first asked about her desire for maintaining coital function
and its importance in her QoL, which will then be integrated into the decision making process. As
an example, if a 75-year-old patient believes that coital function accounts for 10% of her quality
of life, then the new QoL value achieved through the obliterative surgery option will be reduced
by 10% in model calculations.
Given that the initial Pn(to) value is valid for the current age to of the patient, the value of
the preference factor Pn(t) needs to be defined as a decreasing function of age t. According to
Matthias et al. (1997) and Pauls et al. (2007), the desirability for coital function over the general
population is estimated to decrease following a linear pattern at a rate of 1.9% per year, until
the value drops to zero. To this end, we express the preference factor as a function of time, such
that Pn(t) = max{Pn(to)− 0.019(t− to),0} for t ∈ [46,85], where Pn(to) is the current preference
factor value and t > to is the age of the patient. Through this structure, the impact of a patient’s
willingness in preserving coital function can be reflected on the calculation of the earliest age to
recommend obliterative surgery for that patient.
The three types of inputs defined above are used to select a treatment option for a given patient
by considering the potential impacts of these options on improving the QoL of the patient. This
can be achieved by a stochastic characterization of these impacts, which we describe as follows.
Author: Optimal Treatment Policies for Pelvic Organ Prolapse8 Article submitted to ; manuscript no.
2.1.2. Modeling Treatment Options As noted earlier, there are four major categories of
treatment options for a given patient. These treatment categories consist of (1) watchful waiting,
which implies that no treatment action will be taken before the next annual exam; (2) conser-
vative treatments, which correspond to the non-surgical therapy options for POP treatment; (3)
reconstructive surgery, in which the vaginal vaults are supported by surgical procedures; and (4)
obliterative surgery, in which the vaginal opening is closed to prevent prolapse of pelvic organs.
The treatment options above have been studied extensively in the literature in terms of their
impact on a POP patient’s quality of life, where such impacts are assessed using the PFDI and
PFIQ scores. Hence, we utilize such scores in quantifying the transitions between different QoL
levels as a result of a treatment option selected. Clearly, such transitions are probabilistic, which
should be taken into consideration in determining a treatment plan. We use available information
in the clinical literature to characterize these probabilities.
Barber et al. (2006) study the effects of reconstructive surgery and conservative treatments by
comparing PFDI and PFIQ scores before and after these treatment options. The authors conclude
that reconstructive surgeries provide much more significant improvement in PFDI and PFIQ scores
when compared with the outcome after conservative treatments. Similar to that study, Barber et al.
(2007) evaluate and compare the effectiveness of reconstructive and obliterative surgeries using the
same measures and criteria. A noteworthy feature of reconstructive surgery is the role that patient
age plays in the level of QoL improvement after such surgery. To capture such age-dependency in
the transition probabilities, we utilize the findings of Nguyen and Burchette (2008), where patients
from different age groups are studied. As for the watchful waiting option, we refer to Handa et al.
(2004) where the authors provide information on the natural progression and regression rates of
POP over time.
Based on these findings, we characterize the changes in the QoL of a POP patient after each
treatment option according to a normal distribution, where each treatment option has a different
mean and standard deviation. We truncate and discretize these distributions to capture the bounds
and the discrete nature of the QoL levels, repsectively. The mean and standard deviation values for
the treatment options are shown in Table 1, along with the reference studies used in defining these
parameter values. Note that the mean change in QoL level for the reconstructive surgery option is
defined as a function of patient age t, where t varies between 46 and 85. Also, it is assumed that this
mean value is independent of the number of previous reconstructive surgeries and the time since
such a surgery, as each subsequent reconstructive surgery is aimed at restoring the pelvic usually
by fixing a different compartment than the one operated on in the previous surgical procedures
Price et al. (2008).
Author: Optimal Treatment Policies for Pelvic Organ ProlapseArticle submitted to ; manuscript no. 9
The goal of POP treatment planning is to maximize a patient’s expected future quality of life.
Therefore, the returns from choosing a treatment option are measured based on an aggregation of
the QoL levels over the planning horizon. For each year, we consider the QoL level at the beginning
and end of that year, and use the average level as the representative value for QoL in that year.
We also note that each surgical treatment option requires a certain recovery period before there
is an improvement in a patient’s QoL. Similarly, conservative treatments also require frequent office
visits and other treatment related discomfort. These inconveniences are quantified by a disutility
ratio dat which is proportional to the length of the recovery period. The QoL-adjusted reward due
to the improvement of the status of a patient for the year is then discounted by this disutility ratio.
Finally, considering the aging process, the World Health Organization suggests a general discount
factor of 3% when utilizing measures that involve quality of life metrics (Mathers et al. 2008).
Hence, we also utilize such discounting in our model and associated calculations. More specifics on
these measures and their role in the model are provided in Section 2.2 below.
2.2. Model Formulation
In this section we provide the specifics on the model components described in Section 2.1, and
present a complete formulation of an MDP model that can help determine treatment options for
POP patients. As part of our model development, we formally define the following notation:
Decision Epochs. We assume that decisions on treatment options will be made annually for
patients in the age group considered. This is consistent with the results of (Bibbins-Domingo et al.
2017), where the authors recommend an annual pelvic exam for gynecology patients. To this end,
we let t∈ T refer to each decision epoch where T = {46,47, ...,85}.
States. The state of a patient at age t∈ T is denoted by st =< lt, ht >∈ S for all t∈ T , where st is
a two-dimensional vector consisting of the QoL index lt and surgical history ht for the patient. The
domains for these state parameters are defined such that lt ∈ {1,2, ...,10} and ht ∈ {−1,0,1,2,3},
where −1 corresponds to the case of a previous obliterative surgery, while 0,1,2, and 3 correspond
to the number of previous reconstructive surgeries that a patient had.
Actions. The set of treatment options for a patient is defined by A= {Ww,Cs,Re,Ob}, where
Ww is watchful waiting, Cs is conservative treatment, Re is reconstructive surgery, and Ob is
obliterative surgery. The set of allowable actions at a given state st =< lt, ht > would depend on
the surgical history ht of the patient, and is denoted as Aht ⊆A. We provide a list of allowable
actions in each surgical history state ht in Table 3. We note that these allowable actions also apply
to patients with special conditions as discussed previously in Section 2.1, as such conditions may
result in the same surgical restrictions as the corresponding surgical history state.
Author: Optimal Treatment Policies for Pelvic Organ Prolapse10 Article submitted to ; manuscript no.
Transition Probabilities. The transitions between stages are such that for st =< lt, ht > the next
state st+1 =< lt+1, ht+1 > is defined probabilistically as a function of the action at. More specifically,
lt+1 is dependent on a random outcome defined by the treatment option selected, while ht+1 = ht+1
if at =Re, ht+1 = ht for at ∈ {Ww,Cs} and ht+1 =−1 for at =Ob. Probability of transition from
state s to another state s′ under a given treatment a at age t is denoted as pt(st+1 = s′|st = s, at = a).
As discussed in Section 2.1, we utilize clinical data on QoL transitions to build transition prob-
ability matrices for different treatment options in each period. The transitions for each treatment
option a ∈ A are assumed to follow a truncated normal distribution with mean µat and standard
deviation σat where the lower and upper bounds for the distribution are set as 1 and 10, respectively.
We then discretize this truncated normal distribution and define discrete probability distributions
over the state and action space assuming only integer values for QoL levels. The specific numerical
values for these probability distributions are provided in Online Appendix A.
Rewards. The reward of transitioning from a given state s to another state s′ after a selected
treatment option a at age t is denoted by rt(st+1 = s′|st = s, at = a), and is calculated based
on the current QoL, the disutility measure for the option selected, and preference Pn(t) on the
preservation of coital function. Defining s′ =< l′, h′ > and s=< l,h >, we represent this value as
rt(st+1 = s′|st = s, at = a) = 12(l+ l′)(1− dat )(1−Pn(t)), where the QoL value is averaged over the
past and transitioned QoL levels, and discounted by a factor dat ∈ [0,1] representing the disutility
ratio of treatment option a for a patient of age t. The specific values for dat and the recovery period
lengths used for their calculations are shown in Table 2, which are identified through physician
surveys as functions of patient age. Recall regarding the preference over preserving coital function
that if a patient of age to provides her age-dependent preference parameter value as Pn(to), then
for t > to the preference parameter values are defined as Pn(t) = max{Pn(to)− 0.019(t− to),0}.
Based on these definitions, the optimality equations for the POP treatment selection problem
can be stated as follows:
V ∗t (lt, ht) = maxat∈Aht
{∑
<lt+1,ht+1>∈S
pt(lt+1, ht+1|lt, ht, at)1
2(lt + lt+1)(1− datt )(1−Pn(t)) + γE[V ∗t+1(lt+1, ht+1)]}
∀<lt,ht>∈S,t∈T \{T} (1)
where VT (lT , hT ) = 0 and γ is the discount factor based on aging as introduced earlier.
3. Results
3.1. Optimization Based Treatment Policies
In this section we implement our model using the data described in Section 2, and present our
findings on recommended POP treatment options for different patient types. The recommendations
Author: Optimal Treatment Policies for Pelvic Organ ProlapseArticle submitted to ; manuscript no. 11
are aimed at providing guidance to physicians and patients in the selection of a treatment option.
As noted earlier, the planning horizon for our implementation covers an age span between 46-years
and 85-years, and state definitions include 10 distinct QoL levels and 5 different surgical history
measures.
We present the analysis on the optimal POP treatment policies through four distinct cases,
defined by the surgical history and/or relevant restrictions of a patient. In terms of preference over
preserving coital function, the general analysis initially assumes a default Pn(to) value of zero for
a patient of age to. Sensitivity analysis is then performed to examine how the optimal policy is
affected by a patient’s concern for preserving coital function by considering different preference
parameter values.
3.1.1. Case I: All Treatment Options Available We first consider a situation where any
of the four treatment options can be selected for a given POP patient. From a surgical history
perspective, this case corresponds to the history of no more than two reconstructive surgeries
previously performed on a POP patient. In Figure 9a, we present the optimal treatment policy
under this case, where the recommended action for an observed QoL level is shown for each age and
for 0, 1, or 2 previous reconstructive surgeries - as the treatment structure for POP patients with
0, 1 and 2 previous reconstructive surgeries is the same. This similarity in optimal policies under
different surgical history cases is also supported by clinical observations. Price et al. (2008) conduct
an 11-year retrospective study and conclude that most repeat procedures in follow-up reconstructive
surgeries do not involve repairing the same pelvic compartment that was treated in the original
surgery. As a matter of fact, any follow-up reconstructive surgeries actually improve the QoL of
a patient mostly independent of the surgeries conducted earlier. Therefore, such independency
in the effectiveness of surgeries results in the selection of treatment options being insensitive to
reconstructive surgical history of a patient. Also note that the results presented in Figure 9a assume
no preference towards the preservation of coital function, but later in this section we describe how
such preferences impact the optimal policy.
We observe in Figure 9a that for patients with relatively high QoL levels, i.e. at or above 9,
watchful waiting is recommended at all ages. This threshold reduces to 8 between the ages of
68 and 80, and to 7 after age 80. Surgical treatment options are recommended only for patients
with QoL levels below 6 or 7. In general, patients younger than 53 years old should consider
reconstructive surgery if their QoL levels are below 7. Conservative treatment options act mostly
as a transition zone between the surgical options and the watchful waiting option. Between ages
46 and 80, patients of QoL levels 7 or 8 are generally recommended to go through conservative
treatment. Such conservative treatment options are also recommended for patients of very high
age, e.g. 83-85 years old, at even lower QoL levels, as the disutility of going through a surgery
increases significantly at such high ages.
Author: Optimal Treatment Policies for Pelvic Organ Prolapse12 Article submitted to ; manuscript no.
Impact of Preference over Preservation of Coital Function. While the representation in Figure
9a assumes a default Pn(to) value of zero, it is worthwhile to consider the situation when a Pn(to)
value of 0.3 is assumed for the patient. In Figure 9b we present the optimal treatment policy
when preference factor Pn(to) takes value of 0.3. It is observed that conservative treatment is
recommended for patients of a wider range of QoL levels from 6 to 9, specifically between ages
48 and 80. This is expected, because as obliterative surgery becomes undesirable, conservative
treatments create more utility.
The earliest age to recommend obliterative surgery is also affected by the preference over pre-
serving coital function. In Figure 5a, we show the earliest age to recommend obliterative surgery
under different preference parameter Pn(t) values for Case I. The horizontal and vertical axes in
Figure 5a represent a patient’s age t and her preference factor Pn(t) at that age, respectively.
As suggested by the optimal policy under Case I, the earliest age for recommending obliterative
surgery is 74 and thus the results in Figure 5 are displayed for age 74 and onwards only. The QoL
levels in each region in Figure 5a imply that obliterative surgery is recommended for a patient at or
below that QoL level. As observed in Figure 5b, the earliest age to recommend obliterative surgery
generally increases as a patient’s preference over preserving coital function increases but only for
levels above 20%. This increase is almost linear with respect to the preference factor value Pn(t)
when Pn(t) is less than 30%, and then slowly converges at higher Pn(t) levels. Based on this, a
general rule can be concluded as never to recommend obliterative surgery at any age if a patient’s
preference factor for preserving coital function is around 40% or higher. In such situations, other
treatment options should be selected instead.
3.1.2. Case II: No Reconstructive Surgery Option Available The next case we study
is when a patient is unable or unwilling to go through reconstructive surgery. Following the obser-
vation that a maximum of three reconstructive surgeries are allowed for a POP patient, such a
case might also imply that a patient has had three previous reconstructive surgeries in her sur-
gical history. In addition, a POP patient may be unwilling to have reconstructive surgery due to
the associated discomfort or other physical restrictions. Obliterative surgery remains an option in
this case, as it has a high success rate and results in fewer complications and less post-surgery
discomfort.
Since reconstructive surgery is omitted from the potential treatment options, obliterative surgery
becomes the only surgical option to treat a patient. Note that it is rare for a POP patient to
undergo multiple reconstructive surgeries at relative young age, and such a restrictive situation
where a patient is forced to select obliterative surgery at a young age is very unlikely. Hence, the
optimal policies derived under this case are of more value to elder patients. In our analysis, we
Author: Optimal Treatment Policies for Pelvic Organ ProlapseArticle submitted to ; manuscript no. 13
display the results over the full age span from age 46 to 85 but note that the youngest patients going
through obliterative surgery are around age 60 (Fitzgerald et al. 2006). Given these observations,
the optimal policy structure under Case II is shown in Figure 2a, where the blank region corresponds
to impractical ages for obliterative surgery. While the optimal policy might indicate the use of
obliterative surgery at these ages when Pn(t) values of zero are assumed, conservative treatments
are practical in these situations.
As it can be observed in Figure 2a for Pn(to) = 0, obliterative surgery is recommended for
patients with QoL levels below 6 at almost all ages. Similar to the treatment policy illustrated in
Figure 9a for Case I, watchful waiting is recommended for patients with higher QoL levels, while
for conservative treatments there is an increased preference at QoL level 7 for younger patients.
For patients older than 74, the optimal treatment policy stays the same as in Case I.
Impact of Preference over Preservation of Coital Function Preference over preserving coital
function serves as a key factor when recommending an obliterative surgery in Case II. It can be
observed in Figure 2b for Pn(to) = 0.3 that conservative treatments take over obliterative surgery
as the main recommendation for a majority of the patients below QoL level 8, while the earliest
age for recommending obliterative surgery moves to age 81.
The overall impact of the value of the preference factor Pn(t) under Case II is presented in
Figure 3. We display the results starting from age 60 as obliterative surgeries are rarely performed
below such age, which we noted earlier. Compared with the results in Figure 5a, the earliest age
for recommending obliterative surgery is lower for the same Pn(t) values in Figure 3a, due to
the fact that obliterative surgery is the only effective surgical option under Case II. It is also
observed that for patients above the age of 83, the highest QoL level for recommending obliterative
surgery decreases. This is due to the planning horizon terminating at age 85, and for patients
with medium QoL levels the improvement in QoL after obliterative surgery is not sufficient to
counter the discomfort due to the surgery. The earliest age to recommend obliterative surgery also
increases with Pn(t) values as shown in Figure 3b, and converges to age 83 when Pn(t) exceeds
40%. Considering that such a threshold is not significantly different from that of Case I, we again
conclude that the obliterative surgery should not be performed under any case when coital function
accounts for higher than 40% of the QoL for a patient.
3.1.3. Case III: No Obliterative Surgery Option Available A third case we consider is
when only reconstructive surgery is available as a surgical option for the patient. This situation
may be observed in clinical practice due to some patients having reservations about the obliterative
surgery option. Such a case can be interpreted as being equivalent to a case where the preference
on preservation of coital function has a relatively large value, such as 40% or higher.
Author: Optimal Treatment Policies for Pelvic Organ Prolapse14 Article submitted to ; manuscript no.
In Figure 6, we observe just like in the previous case that watchful waiting option is recommended
for patients with high QoL levels, where the threshold of switching to conservative treatment
decreases from QoL level 9 to level 8 at age 64, and further decreases to level 7 at age 80. Recon-
structive surgery is recommended for patients with lower QoL levels at around level 5 between
ages 46 and 79. After age 79, the QoL threshold for switching to conservative treatment gradually
decreases as the effectiveness of reconstructive surgery gets lower and the inconvenience due to the
surgery increases.
To some extent, the optimal treatment plan in Case III resembles Cases I and II. In all these
three cases, watchful waiting, conservative treatment and the surgery options form a three-layer
structure. In this structure, watchful waiting is always the top layer as it is recommended for
patients with high QoL levels, while surgical options form the bottom layer to treat low QoL level
patients. Conservative treatment would serve as the transition zone between watchful waiting and
surgical options for patients with medium QoL levels. However, this transition zone for conservative
treatment in Figure 6 is wider than those in Figures 9a and 2a. The reason is that reconstructive
surgery, as the only surgical option under Case III, has larger disutility and a decreasing effective-
ness as a patient gets older. As a result, conservative treatment is more appropriate for patients
with QoL level 6, and thus it replaces the reconstructive surgery recommendation at that QoL
level.
Note that the preference over preserving coital function is irrelevant under Case III, as oblit-
erative surgery is assumed to be ruled out. Hence, no sensitivity analysis over Pn(t) values is
performed for this case.
3.1.4. Case IV: No Surgical Options Available We also study the case where no surgical
procedures can be considered for a given patient. This case might refer to a surgical history involving
one previous obliterative surgery after which no surgical treatment options can be performed.
Another real life scenario fitting into this case is where a patient can not endure reconstructive
surgery, but also declines to have an obliterative surgery due to the preference of preserving coital
function or for other reasons. In such a situation, the only treatment options available are watchful
waiting and conservative treatments. Similar to the recommended treatment policy structure in
Cases I, II and III, watchful waiting is recommended for patients with higher QoL levels, while the
previous surgery recommendations are now replaced by the conservative treatment option. The
split of these two treatment options in the optimal policy is shown in Figure 4. Starting at age
46, watchful waiting is recommended for POP patients with QoL levels at or above 9, and this
QoL threshold is further reduced to 8 at age 57, and then to 7 at age 80. For all other situations,
conservative treatment options are recommended.
Author: Optimal Treatment Policies for Pelvic Organ ProlapseArticle submitted to ; manuscript no. 15
As a summary for all four cases, in Table 5 we display some key cut-off values for the available
treatment options in terms of patient age and QoL. The cut-off age between watchful waiting and
conservative treatment gradually decreases through Case I to Case IV, from 67 to 56 for QoL level
9, and slightly from 80 to 79 for QoL level 8. As for reconstructive surgery, the recommended cut-off
QoL levels are more gradual for Case III than for Case I. Obliterative surgery cut-off ages and QoL
levels turn out to be the same for Case I and Case II, which apply to patients at very high ages.
The table serves as a general reference summarizing some key points of the optimal policy results
in a simpler format.
3.2. Physician Based Treatment Policies and Value Comparisons
The optimal policies derived in Section 3.1 are based on clinical data that describe the effectiveness
of treatment options for different types of patients. Given this, a relevant question is how well these
policies match the treatment choices currently being recommended by physicians. This is clearly
a challenging question, as such recommendations are based on a physician’s personal views and
experiences which could vary significantly from one physician to the other. In trying to answer that
question, we performed a written survey of obstetricians and gynecologists that are members of the
New England Urogynecologic Society. The survey was aimed towards identifying general insights on
the treatment policies used by gynecologists when treating POP patients of certain ages, QoL level,
and surgical history. The survey questions addressed all four of the cases described in Section 3.1,
capturing the differences in availability of the treatment options. 32 questions were presented, and
in each of the questions physicians were provided with key patient information, which included the
patient’s age, surgical history, and QoL level on a scale of 1 to 10. The respondent was then asked
to choose the treatment option that they would use among the available options for that patient.
The survey received responses from 17 physicians, which were then analyzed to develop a stochastic
representation of physician-based treatment policies used in current practice. The analysis of the
survey data was done as follows. Under each of the four cases, we identified the probability that
the ‘optimal’ treatment option will be selected by physicians, which was based on the distribution
of physician responses to the survey questions related to that case. This allowed us to develop a
probabilistic representation of physician decisions for any given patient. In other words, one can
identify the optimization based treatment policy for a patient, and then probabilistically estimate
the treatment option to be selected for that patient by a physician. These probability distributions
are summarized for each treatment option under each case in Table 4.
With regard to value comparisons, it is easy to calculate the optimization based expected QoL for
a patient of age t and state st =< lt, ht > using the MDP model and the corresponding optimization
results. These simply correspond to V ∗t (st) which are calculated as part of the backward induction
Author: Optimal Treatment Policies for Pelvic Organ Prolapse16 Article submitted to ; manuscript no.
algorithm used to identify the optimal treatment selection decisions. To calculate the expected QoL
under physician based policies for the same patient, we first assume that the selected treatment
option for the patient will be based on the probability distributions shown in Table 4. This structure
implies a randomized Markovian policy, which can be evaluated using a policy evaluation algorithm.
We let qρ(at|st, a∗t ) denote the probability of selecting treatment option at under a physician based
policy, and define uρt (st) as the expected future QoL for a patient of age t in state st. Note that
qρ(at|st, a∗t ) is defined as a function of the optimal action a∗t for the given state st as discussed
in the description of Table 4. We then calculate uρt (st) for all t ∈ T \{T} and st ∈ S through the
following recursion:
uρt (st) =∑at∈A
qρ(at|st, a∗t ){∑
st+1∈S
pt(st+1|st, at)1
2(lt + lt+1)(1− datt )(1−Pn(t)) + γE[uρt+1(st+1)]} (2)
Moreover, we have uρT (sT ) = 0 for all sT ∈ S.
Based on this representation of the physician based policies, we can perform QoL and cost
based analyses on the two types of policies and quantify the value and costs of optimization based
treatment policies over those that are currently being utilized in practice.
A first key question involves as to whether the optimization based policies can help create some
additional value over current policies used by physicians. We answer this question by comparing the
expected QoL levels for patients of different cases under the two policies and calculating the percent
differences in expected QoL when the distribution characteristics of POP patients in the population
are considered. These differences are calculated asV ∗t (st)−u
ρt (st)
uρt (st)
×100% for a given state vector st ∈ S.
For analysis purposes, we first assume that the patients are indifferent towards the preservation
of their coital function when selecting a treatment option, implying the assumption of Pn(to) = 0.
This assumption is later relaxed by showing the results over different Pn(to) values. Moreover, we
also develop in Section 3.2.3 an estimate of the dollar value due to any QoL improvements that
can be achieved by utilizing our optimization results as decision aids.
3.2.1. Value of Optimization Based Treatment Policies for Individuals We first
present in Table 6 the specific values of the percent differences between the optimization and
physician based policies for a given individual patient. For the sake of compactness, in Table 6 we
indicate the improvements in terms of age groups, but more detailed age-specific tables are included
in Online Appendix B. Based on these estimates, we study below the value of optimization over
different initial QoL levels, different patient ages and different coital function preference values.
Value of Optimization over Initial QoL The expected improvements in patients due to opti-
mization based policies for different initial QoL levels are summarized in the bottom rows of the
subtables in Table 6, and visually depicted in Figure 7a. For Cases I, II and III in general, results
suggest that the most significant improvements occur for patients with very high QoL levels. These
Author: Optimal Treatment Policies for Pelvic Organ ProlapseArticle submitted to ; manuscript no. 17
patients typically suffer from lighter discomfort and are easier to treat, and can have their QoL
improved by as much as 20% to 30% by following the optimal treatment policies. The majority of
patients with QoL levels below 8 are also shown to have relatively high QoL improvements ranging
from 10% - 20%. In contrast, for Case IV patients, the expected improvements tend to decrease
over higher initial QoL levels from as high as 40% down to 0%, implying probably the currently
insufficient implementation of conservative treatment options for low QoL patients.
Value of Optimization over Patient Ages For the expected improvements as a function of patient
age, we demonstrate the variations in Figure 7b. It can be observed that Cases I, II, and III show
an increasing improvement trend until around age 65, and then starts dropping afterwards. These
trends suggest that medium aged patients between 55 and 65 years of age are likely to benefit
most from the optimization based policies as a group, and a physician can be more motivated to
refer to the optimization based policies when such a patient is considered. Case IV, on the other
hand, presents a quite different trend when compared with the other cases, where the expected
QoL improvement is relatively low around 8% before age 65 and then increases up to around 18%
at age 80, implying larger potential benefits of implementing the optimization based policies for
elder Case IV patients. The sharp drop at age 85 for Case IV patients is a boundary effect due to
the termination of the planning horizon.
Value of Optimization over Preference Factor Pn(to) As the last set of analysis in this subsection,
we test the physician based treatment policies under different preference factor levels Pn(to) and
compare the expected QoL improvement that can be achieved under different preference situations.
As part of the analysis, we consider Pn(to) values of 0.1, 0.2, 0.3, and 0.4. Since Cases III and IV do
not involve any decisions on obliterative treatment, the comparative analysis is performed for Cases
I and II only. As shown in Figure 8, as a patient’s preference level for preserving coital function
increases, the optimization based treatment policies provide larger improvements in QoL. A likely
reason is that physician based policies may not be placing sufficient emphasis on the preference
over coital function when suggesting obliterative surgeries, thus a larger Pn(to) value would result
in a greater gap between the values yielded by optimization based and physician based policies.
Hence, it can be concluded that the case of no preference, i.e. Pn(to) = 0, which is the assumption
for several of the analyses conducted in this section, implies a lower bound for the expected QoL
improvement due to optimization based policies.
3.2.2. Value of Optimization Based Treatment Policies for the Society While the
expected QoL improvement by utilizing the optimal treatment plan can be estimated for a specific
patient, we are also interested in estimating the overall impact at the societal level within the
U.S. An overall expected QoL improvement value can be calculated by taking into account the
Author: Optimal Treatment Policies for Pelvic Organ Prolapse18 Article submitted to ; manuscript no.
distribution of POP prevalence in the population, for which we adopt the statistical data provided
by Hove et al. (2009) and Wu et al. (2014) on age and severity distributions of POP in the U.S.
The data in these studies suggest the estimates shown in Table 7 for each age group and severity
level. Each entry in the QoL columns in the table corresponds to the estimated percentage of
POP patients in that category for the given age range. A more detailed version of the table with
estimates for specific ages is provided in Online Appendix C.
Given the expected QoL improvements according to the patient’s initial QoL and age as described
earlier in this section, the weighted averages of QoL improvements under each case are then calcu-
lated using the above distribution of POP patients as weights. As shown in Table 6, the expected
QoL improvement in the society due to potential utilization of optimization based policies for the
four different cases are about 18%, 13%, 16%, and 9%, respectively. While the overall expected QoL
improvement in the society based on different case scenarios are quite different from each other,
we realize that the distribution of the four cases among POP patients are not even. Therefore, a
simple average over the four case scenarios will not be representative of the overall improvement
on a societal level. Despite the difficulty of coming up with the exact breakdown of patient pop-
ulations over the four cases due to lack of any such data in the gynecology literature, estimates
were obtained through interviews of gynecologists at a partner hospital. Based on these estimates,
5% of the POP patients belong to Case IV where a patient can not tolerate any of the surgical
options. About 10% of POP patients consider themselves open to all four treatment options includ-
ing obliterative surgery, which corresponds to Case I. Case II, where a patient cannot tolerate
reconstructive surgery but is able to go through obliterative surgery, is somewhat rare in practice,
to the extent that it can be considered negligible at a population scale. Therefore, the remaining
85% of the POP population can be estimated to be Case III patients. Taking into consideration
of the population distribution among the four cases, the overall expected QoL improvement in the
society due to potential utilization of our proposed optimization based policies is estimated to be
15.4%.
3.2.3. Monetary Value Estimates of Improved QoL Given the abstractness of QoL
measures, a natural question involves as to what these percent improvements represent in more
concrete terms. To provide an answer to this question, we express the QoL based improvements
in economic terms, as there exist several studies that develop estimates for quantifying the value
of QoL in dollars. Most mainstream models use a willingness-to-pay structure to measure the
economic value of a quality-adjusted life year. In the case of POP treatment, such quantification is
especially important for the evaluation of social welfare related to the large group of affected female
patients. To this end, based on the aggregated QoL improvements and the overall population of
Author: Optimal Treatment Policies for Pelvic Organ ProlapseArticle submitted to ; manuscript no. 19
POP patients in the U.S., we can compute the expected monetary value of QoL improvements
resulting from potential utilization of optimization based treatment policies for POP.
A list of commonly used QoL value estimates based on willingness-to-pay models is summarized
in the detailed survey paper by Hirth et al. (2000). We consider three major models, and study how
the dollar value of potential QoL improvements can be defined based on the proposed estimates
by each of these models. The three QoL value models we consider are referred to as the human
capital model of Rice and Cooper (1967), revealed preference-safety model of Ghosh et al. (1975),
and contingent valuation model of Jones-Lee et al. (1985). These three models provide different
estimates for the value of a quality-adjusted life year (QALY) under distinct settings. QALY is
a measure of disease burden that considers both the quality and length of life lived by a person.
Each unit of QALY equals one year of living under a perfect health condition. Given the differences
in assumptions and calculation methods, the three models produce different monetary values for
QALY. Hirth et al. (2000) conducts a thorough review of 42 previous studies from the health
economic literature and concludes the median monetary value of a QALY by each of the three
models respectively are $24,777 for the human capital model, $93,402 for the revealed preference-
safety model, and $161,305 for the contingent valuation model. We refer to these three values in
our analyses as low, medium, and high estimates of QALY, and identify financial values under each
estimate level separately.
Similar to the analysis in Section 4.1, we first consider the expected monetary value of optimiza-
tion for an individual patient, and then use these findings to estimate an overall social value in
dollars. To this end, in Table 8 we display the expected dollar value for an individual patient due to
optimization based policies for patients of different age groups and initial QoL levels under medium
QALY estimates. While the entries in the tables are based on medium QALY estimates, the bottom
rows of the subtables contain dollar value estimates under low and high QALY assumptions as well.
We note that of the three QALY assumptions, the low estimates consider the social impact from
the angle of people’s working efficiency, so the corresponding financial value provides an estimate
of improved productivity in the society. The medium estimate relates to a patient’s income, so
the financial impact can be seen to represent the overall monetary value of enhanced welfare. The
high estimate is related to the situation where a patient is considering to have POP symptoms
treated by paying for the medical expenses, and hence the corresponding financial impact provides
an estimate of saved medical costs at a societal level.
As it can be seen in Table 8, the value of potential QoL improvements due to optimal policies
for a given individual patient can be estimated to be around 10 thousand dollars for patients of
Cases I-III. As expected, this value is lower for Case IV patients, i.e. around 3 thousand dollars
based on medium level estimates of QALY. In addition, patients with higher initial QoL levels
Author: Optimal Treatment Policies for Pelvic Organ Prolapse20 Article submitted to ; manuscript no.
appear to have higher expected returns from optimization based policies for all but the last case
of patients. The monetary values can serve as a further motivation for patients and physicians
to adopt optimization based treatment policies, as they provide a more concrete measure of the
potential improvements due to such policies.
In addition to these individual value estimates, we can also obtain the monetary value of opti-
mization on a societal level by considering the population breakdown among the distinct cases as
described in Section 3.2.2. Using this distribution and the national population projections for 2014
by the U.S. Census Bureau, the societal monetary value of optimization based POP treatment
policies over the entire POP population for the three QALY estimate levels are $7.2 billion, $27
billion, and $45 billion, respectively. A lower bound for the annual impact of this value can be
calculated by assuming that the value for each patient will be realized over a 40-year period, i.e.
the planning window considered in this study. In reality, the value realization for most patients
will happen over a shorter period. This implies a potential annual societal value of at least $675
million (based on medium QALY estimates) due to optimization based treatment policies.
3.3. Impact of Optimization Based Treatment Policies on Treatment Costs
While our analysis indicates that there is a significant socio-economic value that can be gained
through optimization based treatment policies for POP, another issue involves as to whether such
policies imply increased costs in POP treatment. More specifically, we investigate whether opti-
mization based policies prefer more costly treatment alternatives when compared with policies that
represent physician recommendations in current practice. This has significance from a trade-off
perspective, but could also be a relevant concern for healthcare providers and insurance companies.
Barber and Maher (2013) estimate the annual cost of treatment for POP related symptoms in 2013
to be around $1.4 billion in the U.S., and forecast a 4% increase in these costs each year. In this
section, we quantify how this overall cost value can change if optimization based policies were to
be followed in treatment selection for POP.
As part of our analysis, we first consider the frequency of the usage of different treatment options
by optimization based and physician based policies. Given that our representation of the physician
based policies is in the form of randomized policies, we estimate the frequency of the usage of
different treatment options by comparing the expected number of times that each treatment action
is performed over the entire population of POP patients, utilizing the age and QoL distribution of
POP patients as weights. These frequencies under each of the four different cases we consider are
shown in Table 9. The third column under each case in the table contains the percent change in the
implementation frequency of each treatment option due to optimization based policies. As it can be
observed, comparing to physician based treatment policies, optimization based treatment policies
Author: Optimal Treatment Policies for Pelvic Organ ProlapseArticle submitted to ; manuscript no. 21
appear to suggest a lower frequency of usage of watchful waiting and conservative treatments for
Cases I, III, and IV. This implies a higher utilization of surgical treatments, with potentially higher
overall treatment costs for the society, as the costs for non-surgical treatments are minimal with
respect to the costs for surgical options. An exception to this is Case III, implying a cost reduction
due to a more frequent usage of watchful waiting under this case.
While the differences in implementation rates can be estimated as such, the cost of a typical
reconstructive or obliterative surgery is somewhat difficult to evaluate. A given surgery can involve
one or more procedures to address POP related symptoms, such as vaginal hysterectomy, abdominal
hysterectomy, cystocele/rectocele repair, vault suspension, and vaginal vault obliteration. In this
study, we utilize the average cost values for reconstructive surgeries as calculated by Calvo et al.
(2017), where the average cost for a reconstructive surgery is estimated to be $5,986. For obliterative
surgery costs, we adopt the costs listed in the Current Procedural Terminology document of the
American Medical Association. The average cost for an obliterative surgery is reported in that
document as being $3,895. The cost estimation for conservative treatments is discussed by Hullfish
et al. (2011) at a level of $1,300, while the cost for watchful waiting is assumed to be zero as such
cost is mostly negligible compared to the other treatment options.
With the frequencies of usage of different treatment methods shown in Table 9, we can develop
estimates for the actual numbers of treatment instances based on the fact that around 200,000
surgeries are performed to treat POP in the U.S annually (Jones et al. 2010). Using this information
in a proportional way, we further show the overall differences in annual treatment costs due to
the implementation of optimization based policies over the entire POP patient population. These
differences are shown in Table 10 for Cases I, III, and IV, as Case II was noted to be seldom
observed in practice. We note that with the increased recommendations for surgical options under
Case I, the overall costs spent on the estimated 10% of the POP patient population is increased by
$19.9 million. Similarly, optimal policies for Case IV use conservative treatments more frequently
than watchful waiting, which results in a slight cost increase of $11.5 million over 5% of the
POP patient population. However, Case III, which corresponds to approximately 85% of the POP
patient population has a reduced frequency of the surgical treatment options under the optimal
policies, resulting in a savings of $223.5 million. In summary, we estimate that the implementation
of optimization based treatment policies can reduce annual treatment costs for POP by around $200
million in the U.S. Hence, optimal policies are expected to result in significant QoL improvements
while at the same time reducing treatment costs.
4. Discussion
As for policy implications, the results and findings of this paper can potentially improve the cur-
rent practice of POP treatments by achieving higher QoL levels for POP patients. To utilize the
Author: Optimal Treatment Policies for Pelvic Organ Prolapse22 Article submitted to ; manuscript no.
optimization based treatment policies, an evaluation of the patient’s QoL is needed by complet-
ing the PFDI and PFIQ questionnaire. Also, a patient would be surveyed about her preference
for preserving coital function to determine its relative importance in quality of life. Based on
these implementations, some policy implications of optimization based treatment policies can be
described as follows.
First, watchful waiting should be the recommended action for most patients with QoL levels
above 7, while surgical treatments should be recommended mostly to patients with QoL levels below
6. However, these cutoff values are not rigid. To that end, conservative treatment methods serve as
a potential option in the transition zone involving QoL levels 5 to 8. In general, for patients with
same QoL levels in this range, the younger patients are better off by utilizing surgical treatments
and more senior patients would benefit from conservative treatments or watchful waiting. As shown
in Table 5, the specifics of these cut-off values vary according to the case considered.
We also demonstrate that there exists a distinct age threshold for choosing obliterative surgery
over reconstructive surgery for a given patient. This cut-off age, specifically when the patient does
not have any preference over preserving coital function, is around 74. In general, a patient younger
than this cut-off age should not go through obliterative surgery. In addition, we show how the
preference of preserving coital function by a patient impacts this cutoff age for choosing obliterative
surgery. In both Case I and II, where obliterative surgery is a potential treatment option, having
some degree of preference for preserving coital function would always move the threshold age up
significantly from 74. A quick reference is that if a patient considers the preservation of coital
function as representing more than 40% of her quality of life, then obliterative surgery should never
be considered as part of the treatment plan.
Through comparison of optimization based and physician based policies we also conclude that
while all POP patients can potentially benefit from using the optimization based treatment policies,
the room for improvement is particularly large for patients younger than 65 years old and with very
low or high initial QoL levels. Our results also indicate that although for some surgical restriction
cases the cost of treatment is slightly increased under optimization based policies, overall at a
societal level the treatment costs are reduced while at the same time improving the expected quality
of life of the patients.
5. Conclusions
With an aging population, POP is becoming an increasingly important disease greatly affecting
the quality of life of females in the society. Therefore, proper treatment planning for POP is of
increasing concern. In this paper, we propose an MDP based model to identify optimal treatment
policies for POP patients in order to maximize their expected future quality of life. We utilize the
Author: Optimal Treatment Policies for Pelvic Organ ProlapseArticle submitted to ; manuscript no. 23
best available data to provide findings which can help physicians to potentially treat POP patients
better in practice.
Our optimization based treatment policies are presented in the form of optimal policy tables,
which can be used by physicians and patients by taking into account several attributes, such as the
patient’s age, QoL according to POP symptoms, preferences over preserving coital function, and
other potential restrictive conditions. While many factors play a role in the selection of a treatment
option, our model aggregates all key factors and provides some general guidelines to serve as aids or
references in making treatment selection decisions. The value of optimization is shown by comparing
optimal treatment polices with practical policies, where the latter are identified through physician
surveys. The results also show significant economic incentives for implementation of the optimal
treatment policies in POP treatment practice, as they were demonstrated to be cost-saving.
As future work on this study, implementation of recommended options as part of actual physician
decisions on a given set of patients can be performed. While this requires a long and significant
clinical effort, any conclusions from such a study can be used to validate the findings or help adjust
model assumptions to reflect any other potentially relevant factors.
References
J. Ahn and J. Hornberger. Involving patients in the cadaveric kidney transplant allocation process: A
decision-theoretic perspective. Management Science, 42(5):629–641, 1996.
O. Alagoz, L. Maillart, A. Schaefer, and M. Roberts. The optimal timing of living-donor liver transplantation.
Management Science, 50(10):1420–1430, 2004.
O. Alagoz, L. Maillart, A. Schaefer, and M. Roberts. Choosing among living-donor and cadaveric livers.
Management Science, 53(11):1702–1715, 2007a.
O. Alagoz, L. Maillart, A. Schaefer, and M. Roberts. Determining the acceptance of cadaveric livers using
an implicit model of the waiting list. Operations Research, 55(1):24–36, 2007b.
O. Alagoz, H. Hsu, A. Schaefer, and M. Roberts. Markov decision processes: A tool for sequential decision
making under uncertainty. Medical Decision Making, 30(4):474–483, 2010.
M. Barber and C. Maher. Epidemiology and outcome assessment of pelvic organ prolapse. International
Urogynecology Journal, 24(11):1783–1790, 2013.
M. Barber, M. Kuchibhatla, C. Pieper, and R. Bump. Psychometric evaluation of 2 comprehensive condition-
specific quality of life instruments for women with pelvic floor disorders. American Journal of Obstetrics
and Gynecology, 185(6):1388–1395, 2001.
M. Barber, M. Walters, and G. Cundiff. Responsiveness of the pelvic floor distress inventory and pelvic
floor impact questionnaire in women undergoing vaginal surgery and pessary treatment for pelvic organ
prolapse. American Journal of Obstetrics and Gynecology, 194(5):1492–1498, 2006.
Author: Optimal Treatment Policies for Pelvic Organ Prolapse24 Article submitted to ; manuscript no.
M. Barber, C. Amundsen, M. Paraiso, A. Weidner, A. Romero, and M. Walters. Quality of life after
surgery for genital prolapse in elderly women: obliterative and reconstructive surgery. International
Urogynecology Journal, 18(7):799–806, 2007.
Matthew D Barber. Pelvic organ prolapse. BMJ, 354:i3853, 2016.
Kirsten Bibbins-Domingo, David C Grossman, Susan J Curry, Michael J Barry, Karina W Davidson, Chyke A
Doubeni, John W Epling, Francisco AR Garcıa, Alex R Kemper, Alex H Krist, et al. Screening
for gynecologic conditions with pelvic examination: Us preventive services task force recommendation
statement. Jama, 317(9):947–953, 2017.
D. Calvo, L. Lopez-Fando, M. Jimenez-Cidre, J. Gomez De Vicente, R. Cabello, I. Laso, and F. Burgos. Cost
analysis of surgical treatment for pelvic organ prolapse by laparoscopic sacrocolpopexy or transvaginal
mesh. Actas Urologicas Espanolas, 42(2):117–122, 2017.
J. Chhatwal, O. Alagoz, and E. Burnside. Optimal breast biopsy decision-making based on mammographic
features and demographic factors. Operations Research, 58(6):1577–1591, 2010.
B. Denton, M. Kurt, N. Shah, S. Bryant, and S. Smith. Optimizing the start time of statin therapy for
patients with diabetes. Medical Decision Making, 29(3):351–367, 2009.
R. Ellerkmann, G. Cundiff, C. Melick, M. Nihira, K. Leffler, and A. Bent. Correlation of symptoms with
location and severity of pelvic organ prolapse. American Journal of Obstetrics and Gynecology, 185
(6):1332–1338, 2001.
H. Falk and S. Kaufman. Partial colpocleisis: The le fort procedure. analysis of 100 cases. Obstetrics &
Gynecology, 5(5):617–627, 1955.
M. Fitzgerald, H. Richter, S. Siddique, P. Thompson, H. Zyczynski, A Weber, et al. Colpocleisis: a review.
International Urogynecology Journal, 17(3):261–271, 2006.
D. Ghosh, D. Lees, and W. Seal. Optimal motorway speed and some valuations of time and life. The
Manchester School, 43(2):134–143, 1975.
R. Gutman, D. Ford, L. Quiroz, S. Shippey, and V. Handa. Is there a pelvic organ prolapse threshold that
predicts pelvic floor symptoms? American Journal of Obstetrics and Gynecology, 199(6):683.e1–683.e7,
2008.
S. Hagen, D. Stark, C. Maher, and E. Adams. Conservative management of pelvic organ prolapse in women.
The Cochrane Database of Systematic Reviews, (4):CD003882, 2006.
V. Handa, E. Garrett, S. Hendrix, E. Gold, and J. Robbins. Progression and remission of pelvic organ
prolapse: a longitudinal study of menopausal women. American Journal of Obstetrics and Gynecology,
190(1):27–32, 2004.
O. Harmanli. Colpocleisis: a simple, effective and underutilized procedure for pelvic organ prolapse. OBG
Management, 21(6):19–26, 2009.
Author: Optimal Treatment Policies for Pelvic Organ ProlapseArticle submitted to ; manuscript no. 25
M. Hauskrecht and H. Fraser. Planning treatment of ischemic heart disease with partially observable markov
decision processes. Artificial Intelligence in Medicine, 18(3):221–244, 2000.
R. Hirth, M. Chernew, E. Miller, A. Fendrick, and W. Weissert. Willingness to pay for a quality-adjusted
life year in search of a standard. Medical Decision Making, 20(3):332–342, 2000.
M. Hove, A. Pool-Goudzwaard, M. Eijkemans, R. Steegers-Theunissen, C. Burger, and M. Vierhout. The
prevalence of pelvic organ prolapse symptoms and signs and their relation with bladder and bowel
disorders in a general female population. International Urogynecology Journal, 20(9):1037–1045, 2009.
K. Hullfish, E. Trowbridge, and G. Stukenborg. Treatment strategies for pelvic organ prolapse: a cost-
effectiveness analysis. International Urogynecology Journal, 22(5):507–515, 2011.
J. Jelovsek, C. Maher, and M. Barber. Pelvic organ prolapse. The Lancet, 369(9566):1027–1038, 2007.
J. Jelovsek, L. Brubaker, and S. Falk. Pelvic organ prolapse in women: Choosing a primary surgical procedure.
Up To Date, 9 Nov. 2015, 2015.
K. Jones and O. Harmanli. Pessary use in pelvic organ prolapse and urinary incontinence. Reviews in
Obstetrics and Gynecology, 3(1):3, 2010.
K. Jones, J. Shepherd, S. Oliphant, L. Wang, C. Bunker, and J. Lowder. Trends in inpatient prolapse
procedures in the united states, 1979–2006. American Journal of Obstetrics and Gynecology, 202(5):
501.e1–501.e7, 2010.
M. Jones-Lee, M. Hammerton, and P. Philips. The value of safety: results of a national sample survey. The
Economic Journal, 95(377):49–72, 1985.
E. Kirkizlar, N. Serban, J. Sisson, J. Swann, C. Barnes, and M. Williams. Evaluation of telemedicine for
screening of diabetic retinopathy in the veterans health administration. Ophthalmology, 120(12):2604–
2610, 2013.
M. Kurt, B. Denton, A. Schaefer, N. Shah, and S. Smith. The structure of optimal statin initiation policies
for patients with type 2 diabetes. IIE Transactions on Healthcare Systems Engineering, 1(1):49–65,
2011.
L. Maillart, J. Ivy, S. Ransom, and K. Diehl. Assessing dynamic breast cancer screening policies. Operations
Research, 56(6):1411–1427, 2008.
J. Mason, D. England, B. Denton, S. Smith, M. Kurt, and N. Shah. Optimizing statin treatment decisions
for diabetes patients in the presence of uncertain future adherence. Medical Decision Making, 32(1):
154–166, 2012.
C. Mathers, D. Fat, and J. Boerma. The Global Burden of Disease: 2004 Update. World Health Organization,
Geneva, Switzerland, 2008.
R. Matthias, J. Lubben, K. Atchison, and S. Schweitzer. Sexual activity and satisfaction among very old
adults: results from a community-dwelling medicare population survey. The Gerontologist, 37(1):6–14,
1997.
Author: Optimal Treatment Policies for Pelvic Organ Prolapse26 Article submitted to ; manuscript no.
A. Montgomery, T. Fahey, Y. Ben-Shlomo, and J. Harding. The influence of absolute cardiovascular risk,
patient utilities, and costs on the decision to treat hypertension: a markov decision analysis. Journal
of hypertension, 21(9):1753–1759, 2003.
J. Nguyen and R. Burchette. Outcome after anterior vaginal prolapse repair: a randomized controlled trial.
Obstetrics & Gynecology, 111(4):891–898, 2008.
I. Nygaard, C. Bradley, D. Brandt, et al. Pelvic organ prolapse in older women: prevalence and risk factors.
Obstetrics & Gynecology, 104(3):489–497, 2004.
A. Olsen, V. Smith, J. Bergstrom, J. Colling, and A. Clark. Epidemiology of surgically managed pelvic organ
prolapse and urinary incontinence. Obstetrics & Gynecology, 89(4):501–506, 1997.
R. Pauls, W. Silva, C. Rooney, S. Siddighi, S. Kleeman, V. Dryfhout, and M. Karram. Sexual function after
vaginal surgery for pelvic organ prolapse and urinary incontinence. American Journal of Obstetrics
and Gynecology, 197(6):1610–1615, 2007.
N. Price, A. Slack, E. Jwarah, and S. Jackson. The incidence of reoperation for surgically treated pelvic
organ prolapse: an 11-year experience. Menopause International, 14(4):145–148, 2008.
E. Ratner, E. Erekson, M. Minkin, and K. Foran-Tuller. Sexual satisfaction in the elderly female population:
A special focus on women with gynecologic pathology. Maturitas, 70(3):210–215, 2011.
D. Rice and B. Cooper. The economic value of human life. American Journal of Public Health and the
Nations Health, 57(11):1954–1966, 1967.
E. Samuelsson, F Arne Victor, G. Tibblin, and K. Svardsudd. Signs of genital prolapse in a swedish population
of women 20 to 59 years of age and possible related factors. American Journal of Obstetrics and
Gynecology, 180(2):299–305, 1999.
A. Schaefer, M. Bailey, S. Shechter, and M. Roberts. Modeling medical treatment using markov decision
processes. In Operations Research and Health Care, pages 593–612. Springer, 2004.
A. Shah, N. Kohli, S. Rajan, and L. Hoyte. The age distribution, rates, and types of surgery for pelvic organ
prolapse in the usa. International Urogynecology Journal, 19(3):421–428, 2008.
S. Shechter, M. Bailey, A. Schaefer, and M. Roberts. The optimal time to initiate hiv therapy under ordered
health states. Operations Research, 56(1):20–33, 2008.
M. Slieker-ten Hove, M. Vierhout, H. Bloembergen, and G. Schoenmaker. Distribution of pelvic organ
prolapse (POP) in the general population; prevalence, severity, etiology and relation with the function
of the pelvic floor muscles. In Neurourology and Urodynamics, volume 23, pages 401–402. Wiley, 2004.
S. Swift, P. Woodman, A. O’Boyle, M. Kahn, M. Valley, D. Bland, W. Wang, and J. Schaffer. Pelvic organ
support study: The distribution, clinical definition, and epidemiologic condition of pelvic organ support
defects. American Journal of Obstetrics and Gynecology, 192(3):795–806, 2005.
Author: Optimal Treatment Policies for Pelvic Organ ProlapseArticle submitted to ; manuscript no. 27
T. Wheeler, H. Richter, K. Burgio, D. Redden, C. Chen, P. Goode, and R. Varner. Regret, satisfaction,
and symptom improvement: Analysis of the impact of partial colpocleisis for the management of severe
pelvic organ prolapse. American Journal of Obstetrics and Gynecology, 193(6):2067–2070, 2005.
J. Wu, C. Vaughan, P. Goode, D. Redden, K. Burgio, H. Richter, and A. Markland. Prevalence and trends
of symptomatic pelvic floor disorders in us women. Obstetrics and Gynecology, 123(1):141, 2014.
Supporting Information
Additional supporting information may be found in the online version of this article:
Appendix A. Transition Probabilities Used in the MDP Model
Appendix B. Expected Percent Improvement in QoL Values over Physician Based Polices by
Implementing Optimization Based Policies
Appendix C. Estimated Distribution of POP Patients in the U.S. by Age Group and QoL
Author: Optimal Treatment Policies for Pelvic Organ ProlapseArticle submitted to ; manuscript no.
Figures
Figure 1 Dynamic treatment decision process for POP.
Prescribe treatment plan Prescribe treatment plan
Observe POP progression
Author: Optimal Treatment Policies for Pelvic Organ ProlapseArticle submitted to ; manuscript no.
Figure 2 Optimal POP treatment policy when reconstructive surgery option is not available (Case II).
(a) Pn(to) = 0.
Age
10
9
8
7
6
5
4
3
2
1
46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85
QoLLevel
Conservative treatments
Watchful waiting
Obliterative surgery
(b) Pn(to) = 0.3.
10
9
8
7
6
5
4
3
2
146 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85
Age
QoLLevel
Conservative treatments
Watchful waiting
Ob
Author: Optimal Treatment Policies for Pelvic Organ ProlapseArticle submitted to ; manuscript no.
Figure 3 Sensitivity analysis around the preference factor Pn(t) under Case II.
(a) Age and QoL level to recommend oblit-
erative surgery based on preference over the
preservation of coital function.
Age
42
41
40
39
38
37
36
35
34
33
32
31
30
29
28
27
26
25
24
23
22
21
20
19
18
17
16
15
62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85
QoL 1
QoL 2
QoL 4
QoL 5
QoL 6
QoL 3
Perform obliterative
surgery if QoL at or
below shown level
Do not recommend
obliterative surgery
Pre
fere
nce facto
r P
n(t
)at tim
e o
f surg
ery
(%
)
(b) Earliest age to recommend obliterative
surgery as a function of preference over the
preservation of coital function.
Preference factor Pn(t) at time of surgery (%) E
arl
iest age t
o r
ecom
mend o
blit
era
tive s
urg
ery
40
45
50
55
60
65
70
75
80
85
0 10 20 30 40 50 60
Author: Optimal Treatment Policies for Pelvic Organ ProlapseArticle submitted to ; manuscript no.
Figure 4 Optimal POP treatment policy when no surgical option is available (Case IV).
Age
10
9
8
7
6
5
4
3
2
1
46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85
Qo
LL
evel
Conservative treatments
Watchful waiting
Author: Optimal Treatment Policies for Pelvic Organ ProlapseArticle submitted to ; manuscript no.
Figure 5 Sensitivity analysis around the preference factor Pn(t) under Case I.
(a) Age and QoL level to recommend oblit-
erative surgery based on preference over the
preservation of coital function.
Pre
fere
nce
fa
cto
r P
n(t
) a
t tim
e o
f su
rge
ry (
%)
Age
38
37
36
35
34
33
32
31
30
29
28
27
26
25
24
23
22
21
20
19
18
17
16
15
14
13
12
11
74 75 76 77 78 79 80 81 82 83 84 85
QoL 1
QoL 2
QoL 3
QoL 4
QoL 5
QoL 6
QoL3
QoL4
Perform obliterative
surgery if QoL at or
below shown level
Do not recommend
obliterative surgery
(b) Earliest age to recommend obliterative
surgery as a function of preference over the
preservation of coital function.
Ea
rlie
st a
ge
to
re
co
mm
en
d o
blit
era
tive
su
rge
ry
Preference factor Pn(t) at time of surgery (%)
72
74
76
78
80
82
84
86
88
0 10 20 30 40 50 60
Author: Optimal Treatment Policies for Pelvic Organ ProlapseArticle submitted to ; manuscript no.
Figure 6 Optimal POP treatment policy when obliterative surgery option is not available (Case III).
Age
10
9
8
7
6
5
4
3
2
1
46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85
Qo
LL
evel
Reconstructive surgery
Conservative treatments
Watchful waiting
Author: Optimal Treatment Policies for Pelvic Organ ProlapseArticle submitted to ; manuscript no.
Figure 7 Expected percent improvement in QoL for different cases.
(a) Expected percent improvement in QoL as a
function of initial QoL.
1086420
40%
30%
20%
10%
0%
Initial QoL Level
Expect
ed Qo
L Imp
rovem
ent
Case ICase IICase IIICase IV
(b) Expected percent improvement in QoL as a
function of patient age.
858075706560555045
20%
18%
15%
13%
10%
8%
5%
Age
Expect
ed Qo
L Imp
rovem
ent
Case ICase IICase IIICase IV
Author: Optimal Treatment Policies for Pelvic Organ ProlapseArticle submitted to ; manuscript no.
Figure 8 Expected percent improvement in QoL as a function of the initial preference factor level for Cases I and
II.
0%
5%
10%
15%
20%
25%
30%
35%
Case I Case III
Exp
ecte
d Q
oL
Imp
rove
me
nt
Pn = 0
Pn = 0.1
Pn = 0.2
Pn = 0.3
Pn = 0.4
Changing Pn_improvementsQoL
Pn(to)=0
Pn(to)=0.1
Pn(to)=0.2
Pn(to)=0.3
Pn(to)=0.4
Author: Optimal Treatment Policies for Pelvic Organ ProlapseArticle submitted to ; manuscript no.
Figure 9 Optimal POP treatment policy when all treatment options are available (Case I).
(a) Pn(to) = 0
10
9
8
7
6
5
4
3
2
1
46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85
Qo
LL
evel
Reconstructive surgery
Conservative treatments
Watchful waiting
Obliterative surgery
Age
(b) Pn(to) = 0.3
Age
Qo
LL
evel
10
9
8
7
6
5
4
3
2
1
46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85
Conservative treatments
Watchful waiting
Reconstructive surgery
Ob
Author: Optimal Treatment Policies for Pelvic Organ ProlapseArticle submitted to ; manuscript no.
Tables
Table 1 The mean and standard deviation of change in the QoL of a patient after each treatment option based
on a normal distribution.
Mean Standard Deviation Reference
Watchful waiting -0.046 0.025 Handa et al. [2004]
Conservative treatment 0.052 0.447 Barber et al. [2006]
Reconstructive surgery 0.286-0.001t 0.371 Barber et al. [2007], Nguyen and Burchette [2008]
Obliterative surgery 0.238 0.540 Barber et al. [2007]
Author: Optimal Treatment Policies for Pelvic Organ ProlapseArticle submitted to ; manuscript no.
Table 2 Disutility of treatment options.
Disutility Factor dat Recovery Period
Watchful waiting (Ww) 0 0Conservative (Cs) 0.020 ∼1 weekReconstructive (Re) 0.0026t− 0.0256 4-10 weeksObliterative (Ob) 0.0009t− 0.0085 1-3 weeks
Author: Optimal Treatment Policies for Pelvic Organ ProlapseArticle submitted to ; manuscript no.
Table 3 Potential treatment options for each surgical history state ht.
Surgical History Indicator ht Potential Treatment Options Aht
ht =−1 A−1 ={Ww,Cs}ht = 0 A0 ={Ww,Cs,Re,Ob}ht = 1 A1 ={Ww,Cs,Re,Ob}ht = 2 A2 ={Ww,Cs,Re,Ob}ht = 3 A3 ={Ww,Cs,Ob}
Author: Optimal Treatment Policies for Pelvic Organ ProlapseArticle submitted to ; manuscript no.
Table 4 Probabilistic representation of the physician based treatment policies as a function of the optimization
based treatment decisions for each case (Ww: Watchful waiting, Cs: Conservative treatment, Re: Reconstructive
surgery, Ob: Obliterative surgery).
(a) Case I: All Treatment Options are Available.
Case I
Probability thatthe physician basedpolicy decision is
Ww Cs Re Ob
When theoptimization basedpolicy decision is
Ww 0.434 0.433 0.133 -Cs 0.675 0.225 0.100 -Re 0.036 0.238 0.655 0.071Ob - 0.347 0.236 0.417
(b) Case II: No Reconstructive Surgery Option.
Case II
Probability thatthe physician basedpolicy decision isWw Cs Ob
When theoptimization basedpolicy decision is
Ww 0.500 0.500 -Cs 0.583 - 0.417Ob - 0.833 0.167
(c) Case III: No Obliterative Surgery Option.
Case III
Probability thatthe physician basedpolicy decision isWw Cs Re
When theoptimization basedpolicy decision is
Ww 0.616 0.267 0.117Cs - 0.667 0.333Re - 0.259 0.741
(d) Case IV: No Surgical Options.
Case IV
Probability thatthe physician basedpolicy decision isWw Cs
When theoptimization basedpolicy decision is
Ww 1 0.5Cs 0.5 0.5
Author: Optimal Treatment Policies for Pelvic Organ ProlapseArticle submitted to ; manuscript no.
Table 5 Summary table showing key cut-off values of patient age and QoL for all four cases considered at
Pn(to) = 0.
Watchful waiting: Apply ifQoL greater than or equal to
Reconstructive Surgery: Applyif QoL less than or equal to
Obliterative Surgery: Applyif QoL less than or equal to
Level 9 Level 8 Level 7 L7 L6 L5 L4 L3 L2 L1 Level 6 Level 5 Level 4
Apply treatmentoption if patientyounger than
Case I 67 80 85 53 73 - - - - 74 83 84 85Case II 66 80 85 - - - - - - - 83 84 85Case III 63 79 85 - 47 79 81 82 83 84 - - -Case IV 56 79 85 - - - - - - - - - -
Author: Optimal Treatment Policies for Pelvic Organ ProlapseArticle submitted to ; manuscript no.
Table 6 Expected percent improvement in QoL due to optimization based policies by age group and symptom
level for different surgical restriction cases.
(a) Case I: All treatment options available.
Age Group QoL 1 QoL 2 QoL 3 QoL 4 QoL 5 QoL 6 QoL 7 QoL 8 QoL 9 QoL 10 Total
46-50 14.1% 14.0% 14.0% 13.9% 13.9% 13.9% 16.4% 15.8% 22.2% 28.7% 16.7%51-55 14.6% 14.5% 14.4% 14.4% 14.4% 14.3% 17.3% 16.1% 23.3% 30.2% 18.2%56-60 15.6% 15.4% 15.3% 15.2% 15.1% 15.1% 17.6% 16.4% 24.6% 31.8% 19.1%61-65 16.6% 16.4% 16.2% 16.1% 16.0% 15.9% 17.2% 16.3% 25.6% 33.1% 19.5%66-70 16.8% 16.5% 16.2% 15.9% 15.7% 15.4% 16.3% 16.3% 25.6% 33.2% 19.4%71-75 14.9% 14.0% 13.2% 12.7% 12.3% 11.8% 13.2% 15.3% 24.1% 30.9% 15.2%76-80 13.2% 11.9% 10.7% 9.6% 8.6% 7.6% 9.0% 13.7% 20.7% 25.8% 11.2%80+ 18.1% 13.8% 11.5% 8.5% 6.8% 4.7% 6.1% 9.9% 13.4% 15.8% 9.4%
Total 16.0% 13.8% 12.8% 12.6% 13.4% 13.5% 15.2% 15.5% 24.1% 31.7% AVG: 17.6%
(b) Case II: No reconstructive surgery option available.
Age Group QoL 1 QoL 2 QoL 3 QoL 4 QoL 5 QoL 6 QoL 7 QoL 8 QoL 9 QoL 10 Total
46-50 12.6% 12.3% 12.2% 12.0% 11.9% 11.4% 13.6% 13.2% 19.9% 26.4% 14.4%51-55 12.9% 12.6% 12.5% 12.2% 12.1% 11.6% 14.5% 13.6% 21.0% 27.8% 15.7%56-60 13.9% 13.6% 13.3% 13.1% 12.9% 12.4% 15.0% 14.1% 22.5% 29.4% 16.8%61-65 15.2% 14.8% 14.5% 14.1% 13.9% 13.9% 14.6% 14.5% 23.9% 31.1% 17.5%66-70 15.2% 14.0% 13.3% 12.5% 11.8% 11.2% 12.5% 14.6% 23.1% 29.5% 14.8%76-80 13.6% 12.2% 11.0% 9.6% 8.6% 7.5% 8.5% 13.2% 19.8% 24.9% 11.2%80+ 18.3% 14.5% 11.2% 8.5% 6.5% 4.6% 5.7% 9.6% 12.6% 15.6% 9.4%
Total 25.2% 20.4% 17.0% 14.4% 13.3% 13.4% 12.8% 13.0% 19.4% 25.4% AVG: 13.2%
(c) Case III: No obliterative surgery option available.
Age Group QoL 1 QoL 2 QoL 3 QoL 4 QoL 5 QoL 6 QoL 7 QoL 8 QoL 9 QoL 10 Total
46-50 12.6% 12.1% 11.6% 11.1% 10.6% 8.8% 8.0% 7.6% 17.5% 23.7% 10.9%51-55 13.0% 12.5% 11.9% 11.3% 10.8% 8.6% 7.8% 7.3% 18.4% 24.6% 11.9%56-60 14.1% 13.4% 12.7% 12.1% 11.5% 9.4% 7.9% 7.4% 19.6% 26.1% 12.9%61-65 15.5% 14.8% 14.0% 13.2% 12.4% 11.8% 8.1% 8.1% 21.0% 27.5% 14.0%66-70 16.7% 15.7% 14.6% 13.8% 12.8% 11.9% 8.3% 12.5% 21.5% 27.8% 15.5%71-75 17.6% 16.0% 14.6% 13.3% 11.9% 10.8% 7.9% 12.5% 20.1% 25.6% 14.2%76-80 20.2% 17.4% 14.6% 12.2% 10.0% 7.9% 5.8% 10.4% 15.9% 20.0% 13.4%80+ 30.4% 21.3% 14.6% 9.8% 6.6% 3.9% 3.1% 6.1% 8.8% 10.6% 12.5%
Total 16.1% 13.8% 12.2% 11.6% 11.9% 11.7% 13.2% 13.8% 22.4% 29.6% AVG: 15.8%
(d) Case IV: No surgical options available.
Age Group QoL 1 QoL 2 QoL 3 QoL 4 QoL 5 QoL 6 QoL 7 QoL 8 QoL 9 QoL 10 Total
46-50 19.9% 17.6% 15.2% 13.2% 11.2% 9.3% 7.5% 5.2% 0.3% 0.0% 7.2%51-55 20.8% 18.4% 15.8% 13.5% 11.2% 9.3% 7.1% 3.6% 0.1% 0.0% 6.4%56-60 22.9% 19.7% 17.0% 14.5% 11.9% 9.6% 7.1% 0.7% 0.1% 0.0% 6.4%61-65 25.8% 21.8% 18.5% 15.4% 12.4% 10.0% 7.2% 0.6% 0.0% 0.0% 7.0%66-70 28.9% 24.7% 20.2% 16.3% 12.9% 10.0% 6.9% 0.4% 0.0% 0.0% 8.6%71-75 33.9% 27.8% 21.7% 16.8% 12.7% 9.2% 5.6% 0.2% 0.0% 0.0% 13.2%76-80 41.4% 31.2% 22.0% 15.9% 10.7% 6.3% 1.6% 0.0% 0.0% 0.0% 18.2%80+ 42.8% 27.3% 16.7% 10.1% 5.5% 2.1% 0.0% 0.0% 0.0% 0.0% 14.0%
Total 38.7% 27.0% 18.8% 14.2% 11.4% 9.0% 6.1% 1.8% 0.1% 0% AVG: 8.6%
Author: Optimal Treatment Policies for Pelvic Organ ProlapseArticle submitted to ; manuscript no.
Table 7 Estimated distribution of POP patients in the U.S. by age group and QoL.
Age Group QoL 1 QoL 2 QoL 3 QoL 4 QoL 5 QoL 6 QoL 7 QoL 8 QoL 9 QoL 10
46-50 0 % 1.2% 3.1% 6.8% 12.4% 17.2% 21.0% 19.5% 12.7% 6.1%51-55 0 % 1.4% 3.4% 6.8% 11.6% 15.5% 18.3% 17.9% 14.4% 10.7%55-60 0 % 1.5% 4.4% 7.9% 12.1% 15.0% 16.6% 16.3% 14.2% 12.0%61-65 0.5 % 2.1% 4.3% 7.8% 12.6% 15.5% 16.5% 15.8% 13.6% 11.3%66-70 4.7 % 4.3% 4.0% 6.8% 12.8% 15.1% 13.8% 13.0% 12.8% 12.7%71-75 10.2% 9.8% 9.5% 10.0% 11.3% 12.1% 12.4% 11.1% 8.3% 5.3%76-80 15.4% 15.1% 14.9% 13.0% 9.5% 8.7% 10.7% 8.9% 3.5% 0.3%80+ 19.1% 18.9% 18.6% 14.9% 7.5% 5.5% 8.7% 6.6% 0.2% 0%
Total 6.2 % 6.8% 7.8% 9.2% 11.2% 13.1% 14.8% 13.6% 10.0% 7.3%
Author: Optimal Treatment Policies for Pelvic Organ ProlapseArticle submitted to ; manuscript no.
Table 8 Estimates of the monetary value (in thousands of U.S dollars) due to optimization based policies for an
individual POP patient.
(a) Case I: All treatment options available.
Age Group QoL 1 QoL 2 QoL 3 QoL 4 QoL 5 QoL 6 QoL 7 QoL 8 QoL 9 QoL 10
46-50 0 2.62 3.92 5.23 6.54 7.75 10.74 11.77 18.68 26.9051-55 0 2.71 4.02 5.42 6.72 8.03 11.21 12.05 19.61 28.2156-60 0 2.90 4.30 5.70 7.10 8.50 11.49 12.24 20.64 29.6161-65 1.59 3.08 4.58 5.98 7.47 8.87 11.21 12.14 21.58 30.9266-70 1.59 3.08 4.48 5.98 7.38 8.59 10.65 12.14 21.48 31.0171-75 1.40 2.52 3.64 4.67 5.79 6.63 8.59 11.49 20.36 29.1476-80 1.21 2.24 2.99 3.55 4.11 4.30 5.98 10.27 17.84 26.6280+ 1.77 2.71 3.27 3.08 3.08 2.43 3.83 7.01 14.10 0
Low Estimate: $3,029 MEDIUM ESTIMATE: $11,420 High Estimate: $19,722
(b) Case II: No reconstructive surgery option available.
Age Group QoL 1 QoL 2 QoL 3 QoL 4 QoL 5 QoL 6 QoL 7 QoL 8 QoL 9 QoL 10
46-50 0 12.61 12.33 11.96 11.58 11.96 12.14 12.33 16.72 21.7651-55 0 12.89 12.52 12.14 11.77 12.70 12.33 12.70 17.28 22.5156-60 0 13.64 13.17 12.61 12.24 12.98 12.80 13.08 18.31 23.4461-65 15.69 14.38 13.73 13.17 12.70 13.08 13.17 12.98 19.24 24.7566-70 16.53 15.69 14.94 14.20 13.54 13.73 13.54 12.52 19.52 25.4171-75 17.65 16.53 15.41 14.38 13.45 12.89 12.80 12.14 18.59 23.1676-80 19.61 17.75 15.88 14.20 12.52 10.74 9.71 10.18 15.32 20.9280+ 33.72 25.22 19.24 14.01 10.46 7.66 4.30 6.44 11.68 0
Low Estimate: $2,719 MEDIUM ESTIMATE: $8,199 High Estimate: $14,160
(c) Case III: No obliterative surgery option available.
Age Group QoL 1 QoL 2 QoL 3 QoL 4 QoL 5 QoL 6 QoL 7 QoL 8 QoL 9 QoL 10
46-50 0 11.58 11.40 11.21 11.11 10.65 12.70 12.33 18.59 24.6651-55 0 11.77 11.68 11.40 11.30 10.83 13.54 12.70 19.61 25.9756-60 0 12.70 12.52 12.33 12.05 11.58 14.01 13.17 20.92 27.4661-65 14.57 13.92 13.45 13.17 12.98 12.98 13.64 13.54 22.42 29.1466-70 14.85 14.38 13.92 13.54 13.17 12.80 13.92 14.10 22.88 29.3371-75 14.10 12.98 12.14 11.58 11.02 10.65 11.68 13.64 21.67 27.8376-80 12.70 11.40 10.18 8.97 8.03 7.10 8.03 12.33 19.05 25.5980+ 18.12 14.10 10.55 7.75 5.79 4.11 5.14 8.59 14.20 0
Low Estimate: $2,175 MEDIUM ESTIMATE: $10,250 High Estimate: $17,701
(d) Case IV: No surgical options available.
Age Group QoL 1 QoL 2 QoL 3 QoL 4 QoL 5 QoL 6 QoL 7 QoL 8 QoL 9 QoL 10
46-50 0 16.44 14.20 12.33 10.46 8.69 7.01 4.86 0.28 051-55 0 17.19 14.76 12.61 10.46 8.69 6.63 3.36 0.09 056-60 0 18.49 15.88 13.54 11.11 8.97 6.63 0.65 0.09 061-65 25.03 20.36 17.19 14.29 11.58 9.34 6.72 0.56 0 066-70 27.37 23.44 19.05 15.22 12.05 9.34 6.44 0.37 0 071-75 31.94 26.06 20.27 15.69 11.86 8.59 5.23 0.19 0 076-80 38.95 29.23 20.46 14.76 9.99 5.98 1.59 0 0 080+ 39.23 24.00 14.48 8.31 4.48 1.59 0 0 0 0
Low Estimate: $899 MEDIUM ESTIMATE: $3,389 High Estimate: $5,853
Author: Optimal Treatment Policies for Pelvic Organ ProlapseArticle submitted to ; manuscript no.
Table 9 Frequency of usage of different treatment options in physician based and optimization based policies.
TreatmentsCase I Case II Case III Case IV
OPT PHY DIFF OPT PHY DIFF OPT PHY DIFF OPT PHY DIFF
Ww 24.2% 30.5% -6.3% 24.5% 33.5% -9.0% 26.0% 16.0% 10.0% 31.1% 65.6% -34.5%Cs 27.6% 29.3% -1.7% 36.3% 44.9% -8.6% 41.0% 42.9% -1.9% 68.9% 34.4% 34.5%Re 38.0% 33.3% 4.7% - - - 33.0% 41.1% -8.1% - - -Ob 10.2% 6.9% 3.3% 39.2% 21.6% 17.4% - - - - - -
Author: Optimal Treatment Policies for Pelvic Organ ProlapseArticle submitted to ; manuscript no.
Table 10 Estimated total annual cost differences (in millions of U.S dollars) of the treatment options in
physician based and optimization based policies under Cases I, III, and IV.
TreatmentCase I Case III Case IV
OPT PHY DIFF OPT PHY DIFF OPT PHY DIFF
Ww - - - - - - - - -Cs 18.4 19.5 -1.1 232.9 242.9 -10.0 23.0 11.5 11.5Re 116.8 102.2 14.6 860.2 1073.7 -213.5 - - -Ob 20.3 13.9 6.4 - - - - - -
Change in cost +$19.9 mil -$223.5 mil +$11.5 mil
Overall difference: -$192.1 mil
Author: Optimal Treatment Policies for Pelvic Organ ProlapseArticle submitted to ; manuscript no.
Appendix A. Transition Probabilities Used in the MDP Model
Table A.1 pt(lt+1 = l′|lt = l, at = Ww).HH
HHHll′
1 2 3 4 5 6 7 8 9 10
1 0.967 0.033 0 0 0 0 0 0 0 0
2 0.015 0.953 0.033 0 0 0 0 0 0 0
3 0 0.015 0.953 0.033 0 0 0 0 0 0
4 0 0 0.015 0.953 0.033 0 0 0 0 0
5 0 0 0 0.015 0.953 0.033 0 0 0 0
6 0 0 0 0 0.015 0.953 0.033 0 0 0
7 0 0 0 0 0 0.015 0.953 0.033 0 0
8 0 0 0 0 0 0 0.015 0.953 0.033 0
9 0 0 0 0 0 0 0 0.015 0.953 0.033
10 0 0 0 0 0 0 0 0 0.015 0.985
Table A.2 pt(lt+1 = l′|lt = l, at = Cs).HH
HHHll′
1 2 3 4 5 6 7 8 9 10
1 0.144 0.148 0.144 0.134 0.118 0.099 0.079 0.060 0.044 0.030
2 0.121 0.130 0.134 0.131 0.121 0.107 0.090 0.072 0.055 0.040
3 0.100 0.113 0.122 0.125 0.122 0.114 0.100 0.084 0.067 0.051
4 0.081 0.097 0.110 0.118 0.121 0.118 0.110 0.097 0.082 0.065
5 0.065 0.081 0.097 0.110 0.118 0.121 0.119 0.110 0.097 0.082
6 0.051 0.067 0.084 0.100 0.113 0.122 0.126 0.123 0.114 0.101
7 0.039 0.054 0.072 0.090 0.107 0.121 0.131 0.134 0.131 0.122
8 0.030 0.043 0.060 0.079 0.099 0.118 0.134 0.144 0.148 0.145
9 0.022 0.034 0.050 0.069 0.090 0.113 0.135 0.153 0.165 0.169
10 0.016 0.026 0.040 0.059 0.081 0.107 0.134 0.160 0.181 0.195
Table A.3 pt(lt+1 = l′|lt = l, at = Re). As the transition probabilities are a function of age t, a sample matrix for
t = 60 is included.HHHHHl
l′1 2 3 4 5 6 7 8 9 10
1 0.104 0.113 0.118 0.119 0.116 0.109 0.099 0.087 0.074 0.061
2 0.090 0.101 0.109 0.114 0.115 0.112 0.106 0.096 0.084 0.072
3 0.078 0.090 0.101 0.109 0.113 0.114 0.111 0.105 0.095 0.084
4 0.066 0.079 0.092 0.102 0.111 0.116 0.117 0.114 0.107 0.097
5 0.056 0.069 0.083 0.096 0.107 0.116 0.121 0.122 0.119 0.112
6 0.047 0.060 0.074 0.089 0.103 0.115 0.124 0.130 0.131 0.128
7 0.039 0.051 0.066 0.082 0.098 0.113 0.127 0.137 0.143 0.144
8 0.032 0.044 0.058 0.074 0.092 0.111 0.128 0.144 0.155 0.162
9 0.026 0.037 0.051 0.067 0.087 0.107 0.129 0.149 0.167 0.180
10 0.021 0.031 0.044 0.061 0.081 0.103 0.128 0.154 0.178 0.199
Author: Optimal Treatment Policies for Pelvic Organ ProlapseArticle submitted to ; manuscript no.
Table A.4 pt(lt+1 = l′|lt = l, at = Ob).HHHHHl
l′1 2 3 4 5 6 7 8 9 10
1 0.103 0.111 0.117 0.118 0.116 0.109 0.100 0.088 0.075 0.062
2 0.089 0.100 0.108 0.113 0.115 0.112 0.106 0.097 0.086 0.073
3 0.076 0.089 0.099 0.108 0.113 0.114 0.112 0.106 0.097 0.086
4 0.065 0.078 0.091 0.102 0.110 0.116 0.117 0.115 0.108 0.099
5 0.055 0.068 0.082 0.095 0.107 0.116 0.121 0.123 0.120 0.114
6 0.046 0.059 0.073 0.088 0.102 0.115 0.125 0.131 0.132 0.129
7 0.038 0.051 0.065 0.081 0.097 0.113 0.127 0.138 0.144 0.146
8 0.031 0.043 0.057 0.074 0.092 0.110 0.128 0.144 0.156 0.164
9 0.026 0.036 0.050 0.067 0.086 0.107 0.129 0.149 0.168 0.182
10 0.021 0.031 0.044 0.060 0.080 0.103 0.128 0.154 0.179 0.201
Author: Optimal Treatment Policies for Pelvic Organ ProlapseArticle submitted to ; manuscript no.
Appendix B. Expected Percent Improvement in QoL Values over PhysicianBased Polices by Implementing Optimization Based Policies
Table B.1 Expected percent improvement in QoL - Case I.
Case I QoL 1 QoL 2 QoL 3 QoL 4 QoL 5 QoL 6 QoL 7 QoL 8 QoL 9 QoL 10
Age 46 14.0% 14.1% 13.9% 13.8% 13.9% 13.9% 14.7% 15.6% 21.8% 28.2%Age 47 13.9% 14.0% 13.9% 13.8% 13.8% 13.8% 16.6% 15.7% 22.0% 28.4%Age 48 14.0% 13.9% 13.9% 13.9% 14.0% 13.9% 16.7% 15.8% 22.2% 28.7%Age 49 14.2% 14.1% 14.1% 14.0% 13.9% 13.8% 16.9% 15.8% 22.3% 28.9%Age 50 14.2% 14.1% 14.1% 14.2% 14.1% 13.9% 17.0% 15.9% 22.6% 29.2%Age 51 14.3% 14.2% 14.1% 14.2% 14.2% 14.0% 17.1% 16.0% 22.8% 29.6%Age 52 14.4% 14.3% 14.3% 14.3% 14.4% 14.2% 17.2% 16.1% 23.0% 29.8%Age 53 14.7% 14.5% 14.4% 14.3% 14.4% 14.1% 17.2% 16.2% 23.3% 30.2%Age 54 14.7% 14.5% 14.5% 14.5% 14.4% 14.4% 17.3% 16.2% 23.6% 30.4%Age 55 14.9% 14.7% 14.7% 14.8% 14.6% 14.6% 17.5% 16.3% 23.8% 30.8%Age 56 15.1% 15.1% 15.0% 14.9% 14.6% 14.8% 17.5% 16.3% 24.1% 31.2%Age 57 15.3% 15.2% 15.1% 15.1% 15.0% 14.9% 17.6% 16.4% 24.4% 31.4%Age 58 15.6% 15.3% 15.3% 15.3% 15.0% 15.0% 17.6% 16.4% 24.5% 31.8%Age 59 15.8% 15.7% 15.6% 15.3% 15.3% 15.3% 17.6% 16.5% 24.8% 32.1%Age 60 16.0% 15.8% 15.8% 15.6% 15.4% 15.5% 17.6% 16.5% 25.0% 32.3%Age 61 16.2% 16.1% 15.9% 15.7% 15.6% 15.5% 17.4% 16.5% 25.3% 32.8%Age 62 16.4% 16.0% 16.0% 15.8% 15.7% 16.0% 17.3% 16.5% 25.5% 32.9%Age 63 16.7% 16.5% 16.3% 16.1% 16.1% 15.8% 17.1% 16.4% 25.8% 33.2%Age 64 17.0% 16.9% 16.7% 16.5% 16.5% 16.1% 16.6% 16.2% 25.9% 33.3%Age 65 16.9% 16.8% 16.3% 16.3% 16.2% 15.8% 17.4% 16.1% 25.8% 33.3%Age 66 16.9% 16.6% 16.5% 16.2% 15.9% 15.6% 17.1% 16.1% 25.8% 33.5%Age 67 16.9% 16.6% 16.5% 16.1% 16.0% 15.7% 16.8% 16.2% 25.8% 33.4%Age 68 16.7% 16.6% 16.3% 16.0% 15.7% 15.4% 16.4% 16.5% 25.7% 33.2%Age 69 16.7% 16.5% 16.1% 15.8% 15.5% 15.2% 15.9% 16.5% 25.5% 33.1%Age 70 16.5% 16.2% 15.8% 15.5% 15.3% 15.0% 15.3% 16.3% 25.2% 32.7%Age 71 16.3% 15.9% 15.4% 15.2% 14.9% 14.5% 14.6% 16.0% 24.8% 32.1%Age 72 15.8% 15.4% 14.8% 14.2% 14.1% 13.7% 13.9% 15.6% 24.5% 31.5%Age 73 14.9% 13.9% 13.3% 12.9% 12.5% 12.3% 13.1% 15.2% 24.2% 30.9%Age 74 14.3% 12.4% 11.4% 10.8% 10.1% 9.4% 12.4% 15.1% 23.6% 30.4%Age 75 13.4% 12.4% 11.3% 10.6% 9.9% 9.0% 11.7% 14.8% 23.2% 29.7%Age 76 13.1% 12.1% 10.9% 10.3% 9.5% 8.6% 10.9% 14.5% 22.5% 28.5%Age 77 13.1% 11.9% 10.7% 9.9% 9.1% 8.2% 10.0% 14.2% 21.8% 27.3%Age 78 13.0% 11.9% 10.6% 9.6% 8.6% 7.7% 9.0% 13.8% 20.7% 25.8%Age 79 13.1% 11.9% 10.5% 9.3% 8.3% 7.1% 8.0% 13.3% 19.9% 24.7%Age 80 13.9% 11.9% 10.7% 9.1% 7.6% 6.5% 7.0% 12.6% 18.7% 22.6%Age 81 14.6% 12.5% 11.1% 8.7% 7.3% 5.8% 6.6% 11.8% 16.8% 19.7%Age 82 16.9% 14.1% 11.3% 8.9% 7.0% 5.0% 6.4% 10.3% 14.2% 16.4%Age 83 21.5% 16.2% 13.3% 9.3% 6.9% 3.7% 5.9% 8.5% 10.6% 12.5%Age 84 23.6% 14.4% 10.9% 6.6% 5.0% 2.4% 4.6% 6.1% 6.9% 7.8%
Author: Optimal Treatment Policies for Pelvic Organ ProlapseArticle submitted to ; manuscript no.
Table B.2 Expected percent improvement in QoL - Case II.
Case II QoL 1 QoL 2 QoL 3 QoL 4 QoL 5 QoL 6 QoL 7 QoL 8 QoL 9 QoL 10
Age 46 12.5% 12.0% 11.5% 11.0% 10.6% 8.9% 8.0% 7.6% 17.3% 23.4%Age 47 12.5% 12.0% 11.5% 11.1% 10.6% 8.8% 8.0% 7.6% 17.4% 23.5%Age 48 12.6% 12.1% 11.6% 11.1% 10.6% 8.7% 8.0% 7.6% 17.5% 23.7%Age 49 12.7% 12.2% 11.6% 11.1% 10.6% 8.7% 7.9% 7.6% 17.6% 23.8%Age 50 12.8% 12.2% 11.7% 11.1% 10.6% 8.8% 8.0% 7.5% 17.7% 24.0%Age 51 12.8% 12.3% 11.8% 11.2% 10.6% 8.7% 7.9% 7.4% 17.9% 24.1%Age 52 12.9% 12.4% 11.9% 11.1% 10.7% 8.6% 7.8% 7.4% 18.1% 24.3%Age 53 12.9% 12.4% 11.8% 11.2% 10.8% 8.6% 7.7% 7.3% 18.3% 24.6%Age 54 13.1% 12.5% 11.9% 11.3% 10.9% 8.6% 7.7% 7.2% 18.6% 24.9%Age 55 13.3% 12.7% 12.1% 11.4% 10.9% 8.6% 7.7% 7.2% 18.9% 25.2%Age 56 13.5% 12.9% 12.3% 11.7% 11.0% 8.8% 7.7% 7.3% 19.0% 25.4%Age 57 13.8% 13.1% 12.5% 11.9% 11.3% 9.0% 7.8% 7.3% 19.3% 25.8%Age 58 14.1% 13.5% 12.7% 12.1% 11.4% 9.2% 7.9% 7.4% 19.5% 26.1%Age 59 14.3% 13.7% 13.0% 12.4% 11.7% 9.5% 7.9% 7.4% 19.8% 26.4%Age 60 14.6% 13.9% 13.2% 12.6% 12.0% 10.3% 8.0% 7.5% 20.1% 26.8%Age 61 14.9% 14.1% 13.5% 12.8% 12.1% 11.4% 8.0% 7.5% 20.4% 27.1%Age 62 15.2% 14.5% 13.8% 13.0% 12.2% 11.6% 8.0% 7.7% 20.7% 27.4%Age 63 15.5% 14.7% 14.0% 13.2% 12.5% 11.7% 8.1% 7.9% 21.1% 27.5%Age 64 15.8% 15.2% 14.4% 13.4% 12.7% 11.9% 8.2% 8.2% 21.3% 27.7%Age 65 16.1% 15.4% 14.5% 13.7% 12.7% 12.2% 8.3% 9.0% 21.5% 27.9%Age 66 16.3% 15.6% 14.7% 13.9% 13.2% 12.2% 8.3% 10.3% 21.6% 28.2%Age 67 16.6% 15.6% 14.5% 13.8% 13.0% 12.1% 8.3% 13.1% 21.6% 28.1%Age 68 16.8% 15.7% 14.5% 13.9% 12.7% 11.9% 8.3% 13.1% 21.6% 28.0%Age 69 17.0% 15.8% 14.8% 13.7% 12.8% 11.8% 8.3% 13.1% 21.4% 27.5%Age 70 17.0% 15.9% 14.6% 13.5% 12.6% 11.6% 8.3% 13.0% 21.2% 27.2%Age 71 17.2% 15.8% 14.7% 13.5% 12.3% 11.3% 8.2% 12.9% 21.0% 26.8%Age 72 17.5% 16.0% 14.7% 13.4% 12.3% 11.2% 8.1% 12.7% 20.7% 26.4%Age 73 17.6% 16.0% 14.6% 13.4% 12.0% 11.0% 8.0% 12.5% 20.3% 25.8%Age 74 17.7% 16.1% 14.6% 13.2% 11.7% 10.5% 7.8% 12.3% 19.6% 25.0%Age 75 18.0% 16.0% 14.3% 13.0% 11.3% 10.0% 7.5% 12.1% 19.1% 24.2%Age 76 18.3% 16.4% 14.3% 12.7% 11.0% 9.3% 7.1% 11.7% 18.1% 23.1%Age 77 18.9% 16.8% 14.2% 12.3% 10.5% 8.6% 6.5% 11.1% 17.2% 21.9%Age 78 19.8% 17.1% 14.5% 12.0% 10.0% 8.0% 5.9% 10.5% 16.0% 20.2%Age 79 21.1% 17.8% 14.7% 12.1% 9.6% 7.3% 5.3% 9.9% 14.7% 18.7%Age 80 23.2% 18.9% 15.3% 11.8% 8.9% 6.3% 4.4% 9.0% 13.4% 16.3%Age 81 25.6% 20.5% 15.6% 11.8% 8.2% 5.1% 3.6% 7.8% 11.6% 14.0%Age 82 30.9% 22.9% 16.1% 11.6% 7.4% 3.7% 3.3% 6.4% 9.3% 11.2%Age 83 37.4% 25.6% 16.8% 10.6% 5.8% 2.0% 2.5% 4.8% 6.6% 7.7%Age 84 34.9% 18.6% 9.2% 3.3% 2.7% 2.3% 1.4% 2.4% 3.3% 3.9%
Author: Optimal Treatment Policies for Pelvic Organ ProlapseArticle submitted to ; manuscript no.
Table B.3 Expected percent improvement in QoL - Case III.
Case III QoL 1 QoL 2 QoL 3 QoL 4 QoL 5 QoL 6 QoL 7 QoL 8 QoL 9 QoL 10
Age 46 12.5% 12.4% 12.2% 12.1% 11.9% 11.5% 11.8% 13.1% 19.6% 26.0%Age 47 12.4% 12.3% 12.2% 11.9% 11.9% 11.3% 14.0% 13.2% 19.7% 26.2%Age 48 12.6% 12.3% 12.2% 11.9% 11.8% 11.4% 14.0% 13.2% 19.9% 26.4%Age 49 12.6% 12.4% 12.2% 12.0% 11.9% 11.5% 14.2% 13.3% 20.1% 26.6%Age 50 12.7% 12.4% 12.3% 12.0% 11.9% 11.5% 14.2% 13.3% 20.3% 26.8%Age 51 12.6% 12.5% 12.3% 12.0% 12.0% 11.5% 14.3% 13.4% 20.5% 27.1%Age 52 12.7% 12.5% 12.3% 12.2% 12.0% 11.6% 14.4% 13.5% 20.7% 27.4%Age 53 12.8% 12.7% 12.5% 12.2% 12.0% 11.7% 14.6% 13.6% 21.0% 27.8%Age 54 13.0% 12.7% 12.6% 12.4% 12.2% 11.6% 14.6% 13.7% 21.3% 28.1%Age 55 13.2% 12.8% 12.8% 12.5% 12.3% 11.7% 14.7% 13.8% 21.5% 28.3%Age 56 13.3% 13.1% 12.8% 12.7% 12.5% 12.0% 14.8% 13.9% 21.9% 28.7%Age 57 13.6% 13.3% 13.1% 12.9% 12.6% 12.2% 14.9% 14.0% 22.1% 29.1%Age 58 13.9% 13.5% 13.3% 13.1% 12.8% 12.3% 15.0% 14.1% 22.4% 29.4%Age 59 14.2% 13.8% 13.5% 13.3% 13.1% 12.7% 15.1% 14.1% 22.8% 29.7%Age 60 14.5% 14.1% 13.8% 13.6% 13.3% 13.0% 15.0% 14.2% 23.1% 30.2%Age 61 14.7% 14.3% 14.0% 13.8% 13.5% 13.3% 14.9% 14.3% 23.4% 30.4%Age 62 15.0% 14.6% 14.2% 13.9% 13.8% 13.5% 14.6% 14.4% 23.7% 30.8%Age 63 15.1% 14.7% 14.4% 14.1% 13.7% 15.0% 14.2% 14.5% 23.9% 31.2%Age 64 15.7% 15.4% 14.8% 14.4% 14.3% 13.9% 13.5% 14.6% 24.3% 31.6%Age 65 15.6% 15.3% 14.9% 14.5% 14.3% 13.9% 15.5% 14.8% 24.4% 31.6%Age 66 15.8% 15.5% 14.9% 14.7% 14.3% 14.0% 15.5% 14.8% 24.5% 31.6%Age 67 15.9% 15.4% 15.0% 14.6% 14.1% 13.9% 15.2% 14.9% 24.6% 31.5%Age 68 15.9% 15.5% 14.8% 14.5% 14.1% 13.7% 15.0% 15.3% 24.5% 31.6%Age 69 16.0% 15.3% 15.0% 14.4% 13.9% 13.6% 14.6% 15.2% 24.5% 31.1%Age 70 15.9% 15.3% 14.8% 14.4% 13.8% 13.3% 14.2% 15.2% 24.4% 30.9%Age 71 16.0% 15.3% 14.8% 13.9% 13.5% 13.1% 13.7% 15.0% 24.1% 30.5%Age 72 15.6% 15.0% 14.4% 13.9% 13.1% 12.8% 13.1% 14.8% 23.6% 30.2%Age 73 15.5% 14.4% 13.7% 13.1% 12.4% 12.0% 12.5% 14.5% 23.1% 29.6%Age 74 15.2% 12.6% 11.8% 10.9% 10.0% 9.4% 11.8% 14.4% 22.7% 29.1%Age 75 13.5% 12.7% 11.6% 10.6% 9.9% 8.9% 11.1% 14.3% 22.2% 28.1%Age 76 13.5% 12.4% 11.4% 10.3% 9.4% 8.7% 10.4% 14.0% 21.8% 27.4%Age 77 13.3% 12.3% 11.2% 9.8% 9.0% 8.1% 9.6% 13.6% 21.0% 26.3%Age 78 13.2% 12.1% 10.8% 9.6% 8.7% 7.4% 8.5% 13.2% 20.1% 25.3%Age 79 13.5% 12.1% 10.8% 9.4% 8.1% 7.0% 7.5% 12.8% 19.1% 23.8%Age 80 14.2% 12.2% 10.7% 9.1% 7.7% 6.2% 6.6% 12.2% 17.2% 21.8%Age 81 15.4% 12.7% 11.0% 9.0% 7.0% 5.8% 6.1% 11.3% 15.2% 19.0%Age 82 17.4% 14.5% 11.5% 8.5% 6.6% 4.9% 5.8% 10.3% 13.0% 16.0%Age 83 20.9% 17.1% 12.2% 9.1% 6.3% 3.6% 5.7% 8.5% 10.3% 12.3%Age 84 23.5% 15.9% 10.4% 6.6% 4.7% 2.4% 4.4% 5.9% 7.2% 8.7%
Author: Optimal Treatment Policies for Pelvic Organ ProlapseArticle submitted to ; manuscript no.
Table B.4 Expected percent improvement in QoL - Case IV.
Case IV QoL 1 QoL 2 QoL 3 QoL 4 QoL 5 QoL 6 QoL 7 QoL 8 QoL 9 QoL 10
Age 46 19.5% 17.2% 14.8% 13.3% 11.1% 9.3% 7.6% 5.4% 0.4% 0.0%Age 47 19.7% 17.2% 15.0% 13.2% 11.2% 9.3% 7.5% 5.3% 0.3% 0.0%Age 48 19.9% 17.6% 15.2% 13.1% 11.2% 9.2% 7.5% 5.2% 0.3% 0.0%Age 49 20.1% 17.8% 15.3% 13.2% 11.3% 9.3% 7.4% 5.1% 0.3% 0.0%Age 50 20.3% 18.0% 15.5% 13.2% 11.3% 9.3% 7.3% 4.9% 0.2% 0.0%Age 51 20.6% 18.0% 15.6% 13.2% 11.2% 9.3% 7.3% 4.7% 0.2% 0.0%Age 52 20.6% 18.1% 15.7% 13.3% 11.2% 9.3% 7.2% 4.3% 0.2% 0.0%Age 53 21.0% 18.4% 15.8% 13.5% 11.2% 9.3% 7.1% 3.8% 0.1% 0.0%Age 54 20.8% 18.8% 15.9% 13.7% 11.3% 9.3% 7.0% 3.0% 0.1% 0.0%Age 55 21.2% 18.9% 16.1% 13.8% 11.3% 9.3% 7.0% 2.2% 0.1% 0.0%Age 56 21.3% 18.9% 16.5% 14.0% 11.6% 9.3% 7.0% 0.8% 0.1% 0.0%Age 57 22.2% 19.5% 16.7% 14.2% 11.8% 9.5% 7.1% 0.7% 0.1% 0.0%Age 58 23.0% 19.5% 16.8% 14.6% 11.9% 9.6% 7.1% 0.7% 0.1% 0.0%Age 59 23.5% 20.2% 17.3% 14.8% 11.9% 9.7% 7.2% 0.7% 0.0% 0.0%Age 60 24.3% 20.5% 17.5% 15.1% 12.2% 9.8% 7.2% 0.7% 0.0% 0.0%Age 61 24.9% 20.8% 18.0% 15.1% 12.2% 9.8% 7.2% 0.6% 0.0% 0.0%Age 62 25.0% 21.4% 18.2% 15.0% 12.3% 10.0% 7.3% 0.6% 0.0% 0.0%Age 63 25.8% 21.9% 18.5% 15.2% 12.4% 9.9% 7.2% 0.6% 0.0% 0.0%Age 64 26.5% 22.0% 18.8% 15.6% 12.5% 10.0% 7.2% 0.5% 0.0% 0.0%Age 65 27.0% 22.6% 19.2% 15.9% 12.7% 10.0% 7.2% 0.5% 0.0% 0.0%Age 66 27.2% 23.5% 19.5% 16.2% 12.8% 10.0% 7.2% 0.5% 0.0% 0.0%Age 67 28.0% 24.3% 19.9% 16.3% 12.9% 10.1% 7.0% 0.4% 0.0% 0.0%Age 68 29.4% 24.6% 20.2% 16.6% 12.9% 10.1% 6.9% 0.4% 0.0% 0.0%Age 69 29.6% 25.3% 20.3% 16.4% 12.8% 9.9% 6.8% 0.3% 0.0% 0.0%Age 70 30.3% 26.1% 21.1% 16.2% 12.9% 9.9% 6.6% 0.3% 0.0% 0.0%Age 71 31.0% 26.8% 21.2% 16.4% 12.9% 9.8% 6.4% 0.3% 0.0% 0.0%Age 72 32.5% 27.1% 21.4% 16.8% 12.9% 9.6% 6.1% 0.2% 0.0% 0.0%Age 73 33.8% 27.7% 21.6% 16.9% 12.8% 9.2% 5.7% 0.2% 0.0% 0.0%Age 74 35.6% 28.0% 22.2% 16.8% 12.6% 8.9% 5.2% 0.1% 0.0% 0.0%Age 75 36.8% 29.2% 21.9% 17.0% 12.4% 8.5% 4.6% 0.1% 0.0% 0.0%Age 76 36.7% 29.9% 22.8% 16.8% 12.1% 7.9% 3.8% 0.0% 0.0% 0.0%Age 77 39.0% 30.9% 22.6% 16.4% 11.5% 7.3% 2.8% 0.0% 0.0% 0.0%Age 78 42.5% 31.5% 22.2% 16.1% 10.9% 6.4% 1.5% 0.0% 0.0% 0.0%Age 79 44.2% 31.3% 21.6% 15.6% 10.1% 5.4% 0.1% 0.0% 0.0% 0.0%Age 80 45.0% 32.5% 20.7% 14.4% 8.8% 4.5% 0.1% 0.0% 0.0% 0.0%Age 81 44.7% 32.2% 19.3% 13.3% 7.4% 3.3% 0.0% 0.0% 0.0% 0.0%Age 82 47.1% 30.6% 18.4% 10.9% 5.5% 2.0% 0.0% 0.0% 0.0% 0.0%Age 83 43.7% 25.5% 15.5% 7.6% 3.7% 0.8% 0.0% 0.0% 0.0% 0.0%Age 84 33.3% 15.6% 9.4% 4.3% 1.9% 0.0% 0.0% 0.0% 0.0% 0.0%
Author: Optimal Treatment Policies for Pelvic Organ ProlapseArticle submitted to ; manuscript no.
Appendix C. Estimated Distribution of POP Patients in the U.S. by AgeGroup and QoL
Table C.1 Estimated distribution of POP patients in the U.S. by age group and QoL.
QoL 1 QoL 2 QoL 3 QoL 4 QoL 5 QoL 6 QoL 7 QoL 8 QoL 9 QoL 10
Age 46 0% 1% 2.9% 7% 13.2% 18.2% 21.7% 19.6% 12% 4.4%Age 47 0% 1.1% 3% 6.9% 12.8% 17.7% 21.3% 19.6% 12.4% 5.2%Age 48 0% 1.2% 3.1% 6.8% 12.4% 17.3% 21% 19.5% 12.7% 6%Age 49 0% 1.4% 3.2% 6.8% 12% 16.6% 20.7% 19.5% 13.1% 6.7%Age 50 0% 1.5% 3.3% 6.7% 11.6% 16.2% 20.3% 19.4% 13.5% 7.5%Age 51 0% 1.6% 3.4% 6.6% 11.2% 15.7% 20% 19.4% 13.8% 8.3Age 52 0% 1.5% 3.4% 6.7% 11.4% 15.7% 19.1% 18.6% 14.1% 9.5%Age 53 0% 1.4% 3.4% 6.8% 11.6% 15.4% 18.3% 17.9% 14.4% 10.8%Age 54 0% 1.3% 3.4% 6.9% 11.8% 15.3% 17.5% 17.2% 14.6% 12%Age 55 0% 1.2% 3.3% 6.9% 12% 15.3% 16.7% 16.5% 14.9% 13.2%Age 56 0% 1.1% 3.3% 7% 12.2% 15.1% 15.9% 15.9% 15.1% 14.4%Age 57 0% 1.3% 3.9% 7.5% 12.1% 15% 16.2% 16.1% 14.7% 13.2%Age 58 0% 1.5% 4.4% 7.9% 12.1% 15% 16.6% 16.3% 14.2% 12%Age 59 0% 1.6% 4.9% 8.4% 12% 15% 17% 16.6% 13.7% 10.8Age 60 0% 1.8% 5.4% 8.8% 12% 14.9% 17.4% 16.8% 13.2% 9.7%Age 61 0% 2% 6% 9.3% 11.9% 14.7% 17.7% 17.1% 12.8% 8.5%Age 62 0% 2% 5.1% 8.6% 12.3% 15.3% 17.1% 16.5% 13.2% 9.9%Age 63 0% 2.1% 4.3% 7.8% 12.7% 15.5% 16.6% 15.9% 13.7% 11.4%Age 64 0.8% 2.1% 3.5% 7.1% 12.9% 15.8% 15.8% 15.2% 14% 12.8%Age 65 1.7% 2.2% 2.6% 6.3% 13.2% 16% 15.1% 14.5% 14.3% 14.1%Age 66 2.6% 2.2% 1.8% 5.5% 13.4% 16.3% 14.4% 13.8% 14.6% 15.4%Age 67 3.6% 3.3% 2.9% 6.2% 13.1% 15.7% 14.1% 13.4% 13.7% 14%Age 68 4.7% 4.3% 4% 6.8% 12.8% 15.2% 13.8% 13% 12.8% 12.6%Age 69 5.8% 5.4% 5.1% 7.4% 12.5% 14.5% 13.5% 12.7% 11.9% 11.2%Age 70 6.9% 6.5% 6.2% 8.1% 12.2% 13.9% 13.2% 12.3% 11% 9.7%Age 71 8% 7.6% 7.3% 8.7% 11.9% 13.2% 13% 11.9% 10.1% 8.3%Age 72 9.1% 8.7% 8.4% 9.3% 11.6% 12.6% 12.7% 11.5% 9.2% 6.9%Age 73 10.2% 9.8% 9.5% 10% 11.3% 12% 12.4% 11.1% 8.3% 5.4%Age 74 11.3% 11% 10.6% 10.6% 11% 11.4% 12.1% 10.7% 7.3% 4%Age 75 12.4% 12.1% 11.8% 11.3% 10.6% 10.8% 11.8% 10.3% 6.4% 2.5%Age 76 13.5% 13.2% 12.9% 11.9% 10.3% 10.3% 11.5% 9.9% 5.5% 1%Age 77 14.6% 14.3% 14% 12.5% 10% 9.5% 11.1% 9.5% 4.5% 0%Age 78 15.5% 15.2% 14.9% 13% 9.5% 8.8% 10.7% 8.9% 3.5% 0%Age 79 16.3% 16.1% 15.8% 13.5% 9.1% 8% 10.2% 8.4% 2.6% 0%Age 80 17.2% 16.9% 16.7% 13.9% 8.6% 7.4% 9.8% 7.9% 1.6% 0%Age 81 18% 17.8% 17.6% 14.4% 8.2% 6.5% 9.4% 7.4% 0.7% 0%Age 82 18.8% 18.6% 18.4% 14.8% 7.8% 5.8% 8.9% 6.9% 0% 0%Age 83 19.4% 19.2% 19% 15% 7.3% 5.2% 8.5% 6.4% 0% 0%Age 84 20% 19.8% 19.6% 15.3% 6.9% 4.5% 8% 5.9% 0% 0%