Download as pdf or txt
Download as pdf or txt
You are on page 1of 19

MODULE 3

RANDOM VABIABLE AND


PROBABILITY DISTRIBUTION
CASE STUDY II (C.8)
 Use of Probability Distribution in Healthcare
Context
 A well known multi – specialty hospital in cardiology located in the south of India provides advanced
cardiac care at an affordable cost to the patients.
 Patients can see doctors at the Outpatient Department (OPD) either by prior appointment (planned) or
by walking in (unplanned).
 The COO of the hospital plans to increase patients’ inflow to generate more revenue.
 Two important aspects: robust planning and allocation, and efficient appointment scheduling
 Therefore, the COO wants:
 an analysis of the footfall data before using a prediction model for patients’ visits
 measure patients’ satisfaction level with OPD services (e.g., waiting time)

 The following metrics (among others) are to be estimated:


 Daily peak period of patients’ visits (10:00 am – 12:00 noon)
 Expected number of patients’ visit during the peak period (300)
 Probability of certain number of patients’ visit in any specified time interval during the peak period
 Average inter-arrival time of patients during the peak period – not required
 Probability of the inter-arrival time being less than 1 min during the peak period – not required
 Probability of patients unsatisfied with the services in OPD during the peak period
PROBABILITY DISTRIBUTION
 Use of Probability Distribution in Healthcare

 X: number of patients’ visit during 10:00 am – 12:00 noon


o X is called a random variable
o X is a function or rule that assigns a numerical value to the outcomes of a sample space
o S: support of X indicates all possible values X can take; X = 0,1,2, …
o The values of X which are called realizations are denoted by 𝑥.

 Probability distribution refers to P(X= 𝑥) for all values X= 𝑥.

 So, we need to model X with a theoretical distribution; note X is a discrete random variable as
it takes only discrete values
EXPECTED VALUE AND VARIANCE
 Use of Probability Distribution in Healthcare
 Like sample frequency distributions, probability distributions also have mean/average and variance
Expected value
X=x P(X=x) Average:
μ = 2.25
0 0.05 μ=𝐸 𝑋
1 0.2 = ෍ 𝑥 𝑃(𝑋 = 𝑥)
2 0.4
Variance: 𝜎2 = 𝑉 𝑋
3 0.2 = ෍(𝑥 − μ)2 𝑃(𝑋 = 𝑥)
4 0.1
= ෍(𝑥)2 𝑃(𝑋 = 𝑥) − (μ)2
5 0.05
SD: 𝜎
𝜎 = 1.17
POISSON DISTRIBUTION
 Use of Probability Distribution in Healthcare
 In this case, a good way to model X is by a Poisson distribution

 When do you model by a Poisson distribution?


o Number of occurrences over a period, length, area or volume or any continuous measure
o Information about average occurrences per unit time, length, area or volume or any continuous measure

 Some examples:
o Number of customers arriving during a time period
o Number of flaws in a square yard of fabric
o Number of bacterial colonies in a cubic centimeter of water
o Number of times a machine fails during a workday
o Number of road accidents per day
o Number of printing mistakes in a book

 Every theoretical distribution is defined by a parameter which imparts uniqueness to the model and characterizes it

𝑒 −𝜆 𝜆𝑥
 P(X= 𝑥)= , 𝑥 = 0, 1, 2, … (This expression is called the probability mass function (pmf) of X)
𝑥!

 𝜆 is the parameter denoting average number of patients’ visit E(X) during 10:00 am – 12:00 noon. This needs to be specified
or estimated before modeling.
APPLICATION OF POSSION DISTRIBUTION
 Use of Probability Distribution in Healthcare
 To answer the following question:
o Probability of the certain number of patients’ visit in any specified time interval
 To model by a Poisson distribution, we need the estimate of 𝜆.
300
 𝜆= = 2.5 is the average number of patients’ visit per minute during 10:00 am –
120
12:00 noon

 What is the probability that 2 patients will visit between 10:30 am – 10:31 am?
o Y: no. of patients’ visit between 10:30 am – 10:31 am
o Average number of patients’ visit between 10:30 am – 10:31 am is 1 ∗ 2.5 = 2.5
o P(Y=2) = POISSON.DIST(2,2.5,FALSE) = 0.2565
o 25% approximately

 What is the probability that 20 patients will visit between 10:30 am – 10:35 am?
o Z: no. of patients’ visit between 10:30 am – 10:35 am
o Average number of patients’ visit between 10:30 am – 10:35 am is 5 ∗ 2.5 = 12.5
o P(Z=20) = POISSON.DIST(20,12.5,FALSE) = 0.0132
o 1% approximately

 What is the probability that at least 10 patients will visit between 10:30 am – 10:35 am?
o P(Z>=10) = 1-POISSON.DIST(9,12.5,TRUE) = 0.7985
o 80% approximately
BINOMIAL DISTRIBUTION – PREMISE
 Use of Probability Distribution in Healthcare
 To answer the following questions:

 Probability of patients unsatisfied with the waiting time in OPD during the peak period

The team working with the COO conducted survey on 200 cardio patients during the peak hour
across various cohorts based on gender, age group, visit type (planned or unplanned) etc.

 Patient’s satisfaction level for waiting time was measured on an ordinal scale rating: Highly
Satisfied (HS), Satisfied (S) and Not Satisfied (NS)

 It was observed that 65% of 200 patients were not satisfied with the waiting time during the
peak hour.

 From this result, we can provide an estimate that there is a 0.65 probability (known as ‘𝒑’)
that a patient will be not satisfied with the waiting time during the peak time in the OPD.
BINOMIAL DISTRIBUTION – MODELING
 Use of Probability Distribution in Healthcare
 Each patient can be either satisfied (HS and S) or not satisfied (NS)

 Outcome for each patient can be treated as binary

 A patient being NS doesn’t affect another patient being NS (independence)

 Probability (𝒑 = 𝟎. 𝟔𝟓) of being NS is same for every patient

 X: number of patients out of ‘𝑛’ who are ‘not satisfied’ with the waiting time during the peak period

 Support of X = 0, 1, … , 𝑛

 X can be modeled by a binomial distribution with number of patients being ‘𝒏’ and probability of
being NS is 𝒑 = 𝟎. 𝟔𝟓

 ‘𝒏’ and ‘𝒑’ are the two parameters of a binomial distribution


APPLICATIONS OF BINOMIAL DISTRIBUTION
 Use of Probability Distribution in Healthcare
 If 20 patients visit, how many of them would not be satisfied on
average?
o Average = 𝑛 ∗ 𝑝 = 20 ∗ 0.65 = 13

 What is the probability that only 10 out of 20 patients will be not


satisfied?
o P(X=10)=BINOM.DIST(10,20,0.65,FALSE)=0.068
o 6.8% chance

 What is the probability that at least 12 out of 20 patients will be not


satisfied?
o P(X>=12)=1-BINOM.DIST(11,20,0.65,TRUE)=0.7623
o 76.23% chance
CONCLUSION
 Use of Probability Distribution in Healthcare
 Hospital has a continuous flow
1. Daily peak period of patients’ visits (10:00 am – 12:00 noon) of patients (3C)
2. Expected number of patients’ visit during the peak period (300)
 Encounters heavy traffic
3. Probability of certain number of patients’ visit in any specified time interval during the peak during the peak period (4, 5)
period
A. What is the probability that 2 patients will visit between 10:30 am – 10:31 am? ≈ 25%  Overall satisfaction related to
B. What is the probability that 20 patients will visit between 10:30 am – 10:35 am? ≈ 1% the waiting time is less (6A,
6B)
C. What is the probability that at least 10 patients will visit between 10:30 am – 10:35 am?
≈ 80%  It is very likely that more than
half patients will not be
4. Average inter-arrival time of patients during the peak period? 0.4 mins = 24 secs satisfied (6D)

 The COO postpones the idea


5. Probability of the inter-arrival time being less than 1 min during the peak period ≈ 92% of increasing patient volume
but wants to solve the problem
6.
of higher waiting times in the
A. Proportion of patients unsatisfied with the services in OPD during the peak period ≈ 65% peak hour.
B. If 20 patients visit during 10:30 – 10:35 am, how many of them would not be satisfied on
average? 𝟏𝟑
C. What is the probability that only 10 out of 20 patients will be not satisfied? ≈ 7%
D. What is the probability that at least 12 out of 20 patients will be not satisfied? ≈
𝟕𝟔. 𝟐𝟑%
NORMAL DISTRIBUTION – PREMISE
 Use of Probability Distribution in Healthcare
 The COO further wants information on

o Average and SD of waiting times of patients during the peak period

o probability that waiting time for a patient will be more than 60 mins

o probability that waiting time for a patient is between 30 and 60 mins

o waiting time corresponding to the 5th percentile

 It was observed that overwhelming majority of the walk-in patients were not satisfied with the
waiting time.

 The team gathered waiting time information for 12000 walk-in patients.
NORMAL DISTRIBUTION – PREMISE
 Use of Probability Distribution in Healthcare
 The histogram of the waiting times
shows a bell shaped and symmetric
curve

 Modeling waiting time (X) with a normal


distribution would make sense

 Two parameters:
 Mean (𝜇)
 SD (𝜎)

 From the collected data:


 Estimate of 𝜇 is 50.4 mins
 Estimate of 𝜎 is 12.6 mins
NORMAL DISTRIBUTION – I

Two parameters Bell Shaped


Mean: 𝜇 = 50.4 Symmetric by mean
SD: 𝜎 = 12.6
68%

95%

99.7% 38
50.4
1𝜎
2𝜎 12.6

3𝜎 Waiting Time
NORMAL DISTRIBUTION – PROB CALC
 Use of Probability Distribution in Healthcare
 probability that waiting time for any patient will be more than 60 mins

 P(X>60)
= 1-NORM.DIST(60,50.4,12.6,TRUE)
= 0.2230

 22.30% chance

Waiting Time
NORMAL DISTRIBUTION – PROB CALC
 Use of Probability Distribution in Healthcare
 probability that waiting time for a patient is between 30 and 60 mins

 P(30<=X<=60)
= NORM.DIST(60,50.4,12.6,TRUE)
− NORM.DIST(30,50.4,12.6,TRUE)
= 0.7242

 72.42% chance

Waiting Time
NORMAL DISTRIBUTION – PROB CALC
 Use of Probability Distribution in Healthcare
 waiting time corresponding to the 5th percentile  P(X<=x)=0.05

 x = NORM.INV(0.05,50.4,12.6)

 x = 29.67

 Only 5% patients wait for less than 30 mins

 These probabilities further justify high level


of dissatisfaction with the waiting times
among the patients who walk-in

 The COO would propose to encourage


patients to visit the hospital by appointment
(rather than walking in) by offering
Waiting Time
discounts, extending OPD timings and
recruiting more health care professionals.
EXPONENTIAL DISTRIBUTION
 Use of Probability Distribution in Healthcare T: Inter-arrival time

 To answer the following questions:


 Average inter-arrival time of patients
 Probability of the inter-arrival time being less than 1 min X1: X2:
Arrival Arrival
 T: time between the arrivals of two consecutive patients time of time of
patient 1 patient 2

 One option to model the probability distribution of T is by an


exponential distribution

 We get an idea from histogram on which distribution to fit

 The probability distribution is defined using a function called


probability density function (pdf).

 The pdf of a random variable following an exponential distribution


 𝑓 𝑡 = 𝜆𝑒 −𝜆𝑡 ; 𝑡 ≥ 0
 𝜆 signifies the rate of arrival (no. of occurrences per unit)
 The value of 𝜆 = 2.5 per minute
EXPONENTIAL DISTRIBUTION
 Use of Probability Distribution in Healthcare T: Inter-arrival time

X1: X2:
 To answer the following questions: Arrival Arrival
 Average inter-arrival time of patients time of time of
 Probability of the inter-arrival time being less than 1 min patient 1 patient 2

 𝜆: average no. of patients arriving per minute

 By unitary method, average of the inter-arrival time Average inter-arrival


E(T) = 1/ 𝜆 = 0.4 mins
time = 1/ 𝝀 = 0.4 mins
 Probability of the inter-arrival time being less than 1 min
o P(T<=1) = EXPON.DIST(1, 2.5, TRUE) =0.92 ≈ 92%

You might also like