Non-state sponsored healthcare economies like India constantly face issues relating to inadequate hospital bed inventory, poor doctor and nurse ratios etc. While supply constraints are well recognised, interestingly enough little emphasis is paid to efficiency in utilization of existing resources. Is it not surprising that large public hospitals run wait times extending into several months for certain surgeries while occupancy rates at such hospitals are less than 80%. If such a situation continues, the queues will keep getting longer and longer especially in an environment when demand for in-patient healthcare is increasing exponentially due to increased awareness, higher incomes and improved payer pools.
While hospitals have done what seemed apparent to ensure smoothening of processes to ensure optimum utilization, the answer is not so simple and such improvements will not yield desired results. The solution clearly lies in harnessing the power of data to make more meaningful process level changes to effect those incremental changes required for achieving optimization.
One of the largest hospitals in Asia, one that is acclaimed for the quality of care and expertise that it possesses, faced a similar problem. While it always seemed that enough and more was being done and that there was a constant rush and that doctors were over-worked, the queue just seemed to get longer and longer. myCOL undertook an extensive analysis of disease level data and produced significant insights around impact on length of stay due to patient attributes, day of week of admission, hospital protocol etc. which resulted in an approximate reduction in length of stay for patients by 8% (in a particular disease). Such reductions not only help improve efficiency at the hospitals but also actually provide great relief in reducing the length of the queue and that too without any additional spends on behalf of the hospital.
Length of Stay (LOS) is the single most important metric to determine efficiency and profitability at a Hospital. It is defined as the number of calendar days from the day of admission to the day of discharge. Most hospitals recognize the importance of LOS and have initiated efforts to improve how data is recorded in this respect. However, analysis is presently restricted to descriptive analytics around average length of stay and how to reduce it. What has been observed empirically is that variances are very high (each case is nuanced) and these variances (statistically speaking long and sometimes fat tails) make ALOS highly misleading.
myCOL undertook analysis of approximately 900,000 patient data at the hospital and using its proprietary models created functions that predict most likely LOS for different disease groups. By inputting some basic variables (including but not limited to Age, Gender, Height, Weight and some other vitals say Blood Pressure for Cardio cases), the myCOL product was able to predict the LOS for each case.
Such accuracy in estimation of LOS in each case (as compared to working with averages) can have significant impact on:
- Accurate estimation of the cost of hospitalization of each individual case thereby potentially increasing the conversion ratio by reducing balking due to high cost
- Availability of the bed, optimal utilization of capacity and capacity planning
- Quality assessment of the department and doctor viz. internal and external benchmarks
- Better customer experience
To illustrate, we analyzed cases for a particular procedure in a particular department at the hospital and observed the following gender based variations. The graph below depicts the Hazard Rate (the probability of discharge upon completion of a particular LOS)
X-axis denotes the LOS
Y-axis denotes the probability of discharge
Stated simply, the graph suggests that the probability of discharge for a given LOS for a female patient in the particular coronary department is lower than that of a male patient. In other words, upon completion of a particular number of days in the hospital the probability that the patient will stay for one more day is higher for female patients as compared to male patients. This single insight has the following impact on the operations:
- Bed occupancy for this department varies based on the gender of the patient. Hence, with the simple use of the gender mix we can match demand to supply with greater accuracy.
- The cost of serving the patient changes with Gender. Hence the estimation at the time of admission will be more accurate than what it is today leading to improved customer experience.
- The hump in the graph indicates that beyond a certain LOS the probability of the patient discharge decreases with each day that the patient spends in the hospital which can help in managing customer expectations along with better inventory management and matching demand and supply.
Detailed analysis was thereafter carried out to determine the impact of Gender on LOS and the following observations were made. All things else being equal
- Women spend 39% more time than men for procedures in the specific department
- For patient admitted on Sunday and operated on Monday, LOS is 64% more than patient admitted on Monday and operated on Tuesday
- For a man admitted on Monday, the median LOS was 8 days
Using the above, the following scenarios were considered
Scenario 1: Woman admitted on Sunday and Man on Monday
From 1 and 3 above, woman median LOS is 11.12 days (8X1.39). So for a Sunday admit, the woman would spend 18.23 days while the LOS for the man (Monday admit) would be 8 days making the aggregate LOS for the Woman and Man to be 26.2368 days
Scenario 2: Man admitted on Sunday and Woman on Monday
The LOS expected for the man is 13.12 days and for the woman it is 11.12 days making the aggregate LOS for the Man and Woman to be 24.24 days.
Basis the above, the Hospital was advised to simply swap the date of admission of the man and the woman (1 day change for either). Such a simple exercise would help the Hospital reduce Patient days by 8% and thereby create additional capacity to that effect without spending anything on capacity building and without making any change in doctor schedules and timings/ length of shifts. Similar exercises were analyzed and implemented for other combinations of patients depending on age, pre-existing disease say diabetes etc. and were found to deliver some very insightful observations which could then be implemented with reasonable success at the Hospital.