Hugo J. Finarelli, Jr.
At a Glance:
Shiftsin demand for a hospital's services can occur unexpectedly. Demand forecasting can help you prepare for these shifts and avoid strategic missteps.
A leading healthcare system spends tens of millions of dollars to build a large heart and vascular center, banking on continued growth in open-heart surgery to generate significant bottom-line profits. A rapid, unexpected decline in open-heart surgery rates makes the financial projections far less favorable.
A community hospital invests heavily in a major emergency department expansion, but builds costly excess capacity because it fails to consider the effects of more efficient patient throughput, the addition of critical care beds, and the creation of an observation unit on the number of treatment stations required.
Each of these scenarios could have been avoided if the health system and the hospital had done a better job of forecasting how changes in treatment patterns would affect future service demands.
Whether an organization is identifying future service opportunities, right-sizing operating capacity, developing a business plan, or developing a capital building project, forecasting the future demand for healthcare services is a critical element in the planning process. Yet it can also be a daunting challenge, especially when changes in technology, clinical practice patterns, competitor initiatives, or payment levels can produce rapid, dramatic shifts in service demand.
There's no denying the uncertainties. But take heart-the factors that drive demand for healthcare services can be identified, and the relationships among these factors can be quantified and projected. A thorough and thoughtful demand forecast is an indispensable means for your organization to better meet community needs and capitalize on opportunities to strengthen the bottom line and financial prospects.
Developing an Accurate Forecast
Successful demand forecasting has two fundamental objectives: to identify the key variables that underlie demand for healthcare services within a particular service area, and to understand how and why these variables might change over time. Accomplishing these objectives requires a systematic analytical process that ensures all aspects of potential demand are assessed. By following a nine-step process, you can create a database and framework for evaluating key variables and testing assumptions, and provide the necessary basis for accurately forecasting demand.
1. Assemble historical data. These data should reflect current and historical demand for the services you wish to examine. It is important to decide at the outset what level of detail you require for the forecast. Inpatient utilization can be analyzed along several dimensions, including patient age, payer, product line (such as cardiology and orthopedics), major service (for instance, medical/surgical, obstetric, and pediatric), or any combination of these dimensions. The data gathered should include both patient activity levels and throughput measures (such as average length of stay, visit duration, or procedure time) if service capacities (including numbers of required beds, operating rooms, or treatment stations) need to be determined.
Other types of data to investigate might include:
- Patient origin and market share by product line and geographic area
- Emergency department (ED) visits by patient type (e.g., admitted, fast-track, psych) and arrival time
- Imaging tests by modality and patient type
- Surgical cases by specialty, patient type, and site
Administrative, financial, and departmental data sometimes vary significantly because each of these areas collects and analyzes different statistics for different reasons. Moreover, many external databases have incomplete, inconsistent, or out-of-date information. It is therefore best to compare several data sources, if possible, to identify the most appropriate set of historical demand data.
2. Analyze historical trends. Examine at least three years of data to identify key trends (absolute change, percentage change, and average annual percentage change) in the services you wish to include in the forecast. Developing ratios between measures of demand may also be helpful (for example, admissions through the ED versus number of ED visits; inpatient imaging tests [by modality] per medical/surgical admission; and CT scans versus MRIs). Big swings in such ratios over a short time are unusual and may signal an underlying data problem.
Changing data classification schemes should also be noted. For example, defibrillator implants were included in the DRGs for valve surgery (104 and 105) prior to October 2001, but were assigned their own DRGs after that date. If you didn't take into account the DRG change, you would almost certainly overstate the decline in open-heart surgery volume that most hospitals experienced between 2000 and 2003.
3. Identify key demand drivers. Key drivers of population-based demand include population growth and aging and changes in technology or treatment patterns that affect service-specific use rates. Other demand factors specific to a particular service, such as the substitution of noninvasive procedures for certain types of surgery, may need to be identified and incorporated into the analysis. Other drivers affect provider-level demand once total market demand has been determined. (See "Population-Based Demand Forecasting" on page 54 and "Provider-Level Demand Forecasting" on page 56.)
Causal and influencing factors should be carefully reviewed and discussed to ensure that all key demand drivers have been considered and that the relationships between the drivers and service demands are well understood.
4. Identify relevant benchmarks. Such benchmarks provide a point of reference for determining the extent to which demand trends in your service area are in line with broader market-place or national trends. Relevant benchmarks include use rates in comparable markets, established best practices or treatment protocols, or service-specific guidelines and performance measures.
Potential sources include:
- National or statewide databases
- Established best practices and protocols
- Guidelines and performance measures published by organizations like the American College of Cardiology and the American College of Surgeons
- Medical journals that report results of clinical trials
- Web sites of recognized market leaders
Without relevant benchmarks, you could easily assume that recent trends in your service area or within your organization will continue for the foreseeable future, when actually current practice patterns may be poised to shift toward norms reported elsewhere. You may therefore find it worthwhile to investigate alternate scenarios, such as a status quo trend based on your organization's unique circumstances versus a shifting trend based on circumstances affecting the entire marketplace or nation.
5. Model existing conditions. Develop a spreadsheet model that best replicates the latest verifiable market data and utilization statistics, and that best projects the trends that have occurred since. One difficulty is that the latest verifiable market data or utilization statistics for other providers may be more than a year old. Nevertheless, historical data and historical trends should be used to develop the most reasonable combination of assumptions about current conditions for the key demand drivers. If the model cannot replicate existing conditions, it cannot be used to predict future demand.
For instance, ED visits have three key demand drivers: service area population growth, ED visits per 1,000 population in the service area, and the provider's ED market share within the service area. It's important to develop reasonable assumptions for each of these drivers, especially if a hospital has experienced several consecutive year-to-year increases of 5 percent or more in ED demand.
6. Develop core assumptions for population-based demand. Key factors affecting such demand include population growth, aging, and use rates (see "Population-Based Demand Forecasting" on page 54). These factors are often called external factors because they are outside of the healthcare organization's control. Information for making assumptions about population growth and aging generally are available from demographic firms or from national, state, or local governmental agencies. Core assumptions regarding population-based use rates usually must be developed on a case-by-case, market-by-market basis. The core assumptions should take into account historical trends, external benchmarks, rates experienced elsewhere, the expected effect of technology or medical advances on treatment patterns or location of care, anticipated effects of increases or decreases in uninsured population, and potential changes in other related factors. For each service you are considering, combining the core use-rate assumptions with the population projections provides a forecast of total market demand.
7. Develop core assumptions for provider-level demand. Factors that determine demand at the provider level include market share, patient mix or flow patterns, and operational performance. (see "Provider-Level Demand Forecasting"). These factors are often called controllable factors because they can be affected by specific actions of the provider.
For instance, targeted market-share gains can be linked to strategic/new service initiatives. Patient mix or flow can be changed through measures such as aggressively identifying observation patients for treatment and discharge, setting stricter criteria for admission to critical care units, and developing step-down units to encourage use of less resource-intensive settings when medically appropriate. And targeted performance improvements can be achieved by reducing length of stay or acquiring new technologies that can improve turnaround time and productivity (for example, a high-speed CT scanner or ED patient tracking system). All such initiatives should be considered in developing the core assumptions for departmental workload levels.
When developing market-share assumptions, don't overlook the potential impact of new competitors in the market place, including entrepreneurial physicians or physician groups and national niche providers. Many hospitals have lost significant market share because of new competitors.
8. Create a baseline forecast of future demand. This forecast should combine the core assumptions for both population-based and provider-level demand. Typically, a baseline forecast includes:
- Moderate assumptions for external factors to create the baseline forecast of population-based demand
- Moderate market-share targets to create the baseline forecast of provider-specific utilization levels
- Aggressive performance-improvement targets to develop the baseline workload projections
Using the most aggressive performance targets is considered good business planning and tends to moderate the increase in resource requirements (e.g., staffing levels, facility capacity) that might otherwise accompany projected increases in inpatient volumes.
Projections should be developed for three to five years in the future. For services that may be affected by dramatic changes in demand drivers, forecasts will need to be revisited relatively frequently. On the other hand, a three- to five-year planning horizon is too short for major building projects or new hospital construction projects with long life cycles. For such projects, it's best to assume very moderate changes in most key drivers beyond the three- to five-year horizon.
9. Test sensitivity of projections to changes in core assumptions. Consider alternative scenarios with different sets of assumptions. Such scenarios might include:
- Low and high rates of change in population-based use rates
- More dramatic shifts in market share (aggressive growth versus loss of market share)
- Results that, for whatever reason, fall short of achieving projected operational efficiencies or other performance-improvement targets
It is useful to consider a best-case scenario, with more favorable assumptions (such as higher population-based demand, greater market share growth) than were used in the baseline forecast.
It's probably more important, however, to test downside sensitivity by using less favorable (or more conservative) assumptions about use rates, market share, or performance improvement. "Go" or "no-go" decisions on certain investments or proposed capital projects may hinge on the risk associated with not achieving targeted utilization levels.
Sensitivity analysis is not intended to introduce so much uncertainty that decision makers are afraid to act. Instead, it provides a reasonable estimate of the risk and reward associated with a strategic initiative (to build, expand, or invest) by considering realistic combinations of internal and external factors that might cause future demand to diverge from the baseline forecast.
An Art and a Science
Given the rapidly changing healthcare environment, projecting demand with a reasonable degree of certainty may seem unachievable. However, the steps outlined above provide a solid framework for systematically and thoughtfully tracking and explaining potential future changes in demand, both expected and unexpected. Finally, remember demand forecasting is an art as well as a science. Be creative!
Publication Date: Monday, November 01, 2004