What Is Health Care Utilization and How Is It Measured?

Health care utilization refers to the quantity and pattern of health services that people actually use, from a routine doctor visit to an emergency department trip to a month-long prescription. Measuring it sounds straightforward, but in practice it involves stitching together insurance claims, hospital records, patient surveys, and electronic health records, each with its own blind spots. The field has developed a rich toolkit for tracking who uses what services, how often, and why, and the answers shape everything from insurance premiums to hospital staffing to national health policy.

What Counts as Health Care Utilization

At its broadest, health care utilization covers any interaction between a person and the health care system. Researchers and policymakers typically break it into several categories: inpatient care (hospital admissions, length of stay, readmissions), outpatient care (primary care visits, specialist visits, same-day procedures), emergency department visits, preventive services (screenings, immunizations, annual check-ups), pharmaceutical use (prescriptions filled, days of medication supply), behavioral health services (therapy sessions, psychiatric visits, substance use treatment), and post-acute or long-term care (skilled nursing stays, home health visits, rehabilitation). Each category has its own measurement conventions and quirks, and lumping them together obscures more than it reveals.

The distinction between a “visit” and an “episode” matters here. A single visit is one encounter on one day. An episode of care might span weeks and include an initial diagnosis visit, lab work, a follow-up appointment, and a filled prescription, all counted separately but tied to one underlying health problem. Depending on the question being asked, researchers might count visits, episodes, total days of care, or spending. The choice changes the story the data tells.

Why People Use Health Services in the First Place

The most widely used framework for understanding utilization patterns comes from Ronald Andersen, whose behavioral model sorts the drivers of health care use into three buckets. Predisposing factors are demographic and social characteristics like age, gender, education, marital status, and employment. Enabling factors are the resources and logistics that make care accessible, including income, insurance coverage, where you live, and whether providers are nearby. Need factors are the health conditions themselves: chronic illnesses, acute symptoms, and how sick someone perceives themselves to be.

A systematic review of studies applying Andersen’s model found that most research examines age, education, gender, marital status, and employment as predisposing factors, while income, insurance, and location dominate the enabling category. Chronic illness and self-rated health are the most common need factors studied.¹ This three-part framework is useful because it reminds you that utilization is never just about how sick people are. Two people with the same diagnosis can have wildly different patterns of care depending on their insurance, transportation, and proximity to a clinic.

The Three Main Data Sources

Most utilization measurement relies on one of three data streams: administrative claims, electronic health records, and self-reported surveys. Each has strengths that compensate for the others’ weaknesses, and researchers frequently cross-check them to figure out what is actually happening.

Administrative Claims

Insurance claims data is the workhorse of utilization measurement. Every time a provider bills an insurer, a record is generated that captures the patient, the diagnosis code, the procedure, and the cost. Claims cover enormous populations and require no extra effort from patients or clinicians. They are especially good at tracking services with unique billing codes. One study comparing claims against medical records found that agreement was high for treatments and tests with distinct billing codes, like gene expression profiling and trastuzumab, but much lower for tests without unique codes.² German hospital claims data has similarly been shown to capture information useful for quality-of-care assessments, including inpatient management patterns and in-hospital mortality.³

The catch is that claims only record what gets billed. Services provided for free, care received outside the insured network, and anything that a provider does not submit a claim for simply vanishes from the data. Claims also tell you what was billed, not necessarily what was clinically appropriate. And in some insurance systems, encounter data can be incomplete. A study of Medicare Advantage encounter records for post-acute care found that only about 82% of skilled nursing facility assessments and 63% of home health assessments could be matched to corresponding encounter records in 2019.⁴

Electronic Health Records

EHRs capture clinical detail that claims miss: lab results, vital signs, clinical notes, and referrals. For preventive services performed in a primary care setting, like blood pressure measurement and BMI recording, automated EHR extraction shows near-perfect agreement with manual chart review. But for services commonly referred out, like cancer screenings, agreement drops. Breast cancer screening showed only moderate agreement between automated extraction and chart review, while colorectal cancer screening was somewhat better.⁵ The problem is that when a patient gets a mammogram at an outside facility, the result may sit in a fax or a scanned document rather than a structured data field, making it harder for automated tools to find.

The adoption of EHRs has also affected utilization itself, not just its measurement. One study found that practices achieving meaningful use of EHRs had about 17 fewer primary care visits and 61 fewer lab tests per 100 patients compared to practices using EHRs without meeting meaningful-use criteria.⁶ The implication is that how a health system organizes its digital tools can change how much care patients receive, which complicates any before-and-after comparison.

Self-Reported Surveys

National health surveys ask people directly about their care. This captures services that claims and EHRs miss, like visits to out-of-network providers, traditional healers, or alternative medicine practitioners. But people misremember. A study tracking rehabilitation patients found that self-reported GP visits under-reported actual visits by about 14% over a full year, and the pattern was uneven: patients over-reported visits by roughly 35% in the first six months after discharge and under-reported by about 36% in the second six months.⁷ Another study comparing self-reports to administrative records found that agreement ranged from 30% to 99% depending on the service, with annual doctor visits showing the lowest agreement. Younger, more educated, healthier males tended to report more accurately.⁸

The question wording itself matters, especially for behavioral health. A study comparing different survey question sets for mental health service use found that estimates varied substantially depending on how the questions were phrased and where they appeared in the survey. Agreement across question sets ranged from marginal to good, and the characteristics of identified service users shifted depending on which questions were asked.⁹

Hospital and Emergency Department Metrics

For inpatient care, the most tracked numbers are admission rates, average length of stay, and readmission rates. Length of stay has been declining for decades. A 14-year study of Veterans Affairs hospitals found that risk-adjusted mean length of stay dropped from about 5.4 days to about 4 days, while 30-day readmission rates also fell, from roughly 16.5% to 13.8%.¹⁰ This matters because a common worry is that shorter stays push patients out too early, causing readmissions. But an international study found that readmitted patients actually tended to have had longer initial stays, not shorter ones, suggesting readmissions were driven more by disease severity than premature discharge.¹¹

Emergency departments present their own measurement challenge. “Frequent users” are a small share of patients who account for a disproportionate share of visits. One study at a regional trauma center found that about 71% of enrolled patients had two or more ED visits within the prior year, roughly half had three or more, and about 15% had five or more.¹² Among older adults, the probability of being a frequent ED user varied by visit type: it was lowest for minor injuries and highest for visits related to alcohol, drug, and psychological issues.¹³ Cancer patients made up about 18% of frequent ED users in one large dataset, with a particularly high concentration in urban emergency departments.¹⁴

Measuring Medication Use

Pharmaceutical utilization has its own measurement ecosystem. The World Health Organization developed the Anatomical Therapeutic Chemical classification and Defined Daily Dose (ATC/DDD) system as a standard way to compare drug consumption across populations and over time.¹⁵ The DDD represents the assumed average maintenance dose per day for a drug’s main indication. It gives you a standard measuring stick, but it does not always match real-world prescribing. A study comparing DDD estimates to actual days’ supply recorded in US pharmacy claims found that concordance depended heavily on the drug class.¹⁶

For adherence, the most common claims-based measure is the proportion of days covered (PDC), which calculates the fraction of days in a period that a patient had medication available. But assumptions about the prescribed daily dose have an outsized influence. A study of heart failure patients found that varying the assumed dose swung PDC estimates from about 42% to 84% for the same patients.¹⁷ That is an enormous range and a reminder that a single adherence number without sensitivity analysis can be misleading.

Preventive Care and the Pandemic Disruption

Utilization of preventive services is measured slightly differently because the question is not just “did people use this?” but “did the right people get the right screening at the right time?” Researchers track rates of recommended screenings, like blood pressure checks, cholesterol testing, and cancer screenings, and compare them against guideline-recommended intervals.

The COVID-19 pandemic created a natural experiment in how quickly preventive utilization can collapse. By 2021, screening rates in the United States still had not returned to pre-pandemic levels. Eligible adults were about 5% less likely to have had blood pressure and blood glucose screenings, 7% less likely to have had cholesterol checked, and 12-14% less likely to have received cervical or prostate cancer screenings compared to 2019.¹⁸ The drop was not uniform across racial and ethnic groups, which highlights that measuring the average can hide important disparities.

How Geography and Cost Sharing Drive Variation

Utilization varies enormously from one region to another, and understanding why is one of the field’s central puzzles. A study of geographic variation in health spending across small areas found that demand-side factors, including local income, urbanization, and demographics, explained the vast majority (about 87%) of the overall variation. On the supply side, the number of hospital beds per capita was a strong predictor: each additional bed per thousand people was associated with a 2.9% increase in per-capita inpatient spending.¹⁹ The implication is that what drives utilization is partly need, partly ability to pay, and partly the sheer availability of beds and providers.

Cost sharing, meaning the portion of a bill the patient pays, is a powerful lever on utilization. Research exploiting a sharp reduction in cost sharing at age 70 in Japan found that both outpatient and inpatient care were price sensitive among older adults: when costs dropped, use went up.²⁰ In the US Medicare population, a study of retired public employees whose supplemental insurer raised cost sharing found that doctor visits and prescriptions declined, as expected, but hospital admissions increased, an “offset” effect where skipping cheaper outpatient care led to costlier inpatient care down the road.²¹ A broader systematic review confirmed this pattern: higher cost sharing tended to decrease outpatient visits but increase inpatient use.²² This is one of the clearest cases where looking at a single utilization metric in isolation gives you a dangerously incomplete picture.

Measuring Overuse and Underuse

Utilization measurement is not just about counting services. It is also about evaluating whether the right services are being delivered. Overuse, sometimes called low-value care, refers to services that provide little or no benefit relative to their cost or risk. Using Medicare claims, researchers developed indicators for low-value services and found that, depending on how tightly the indicators were defined, such services affected between roughly a quarter and 41% of beneficiaries, accounting for somewhere between 0.6% and 2.7% of annual spending.²³ Building measurable indicators from clinical recommendations like the Choosing Wisely lists turns out to be difficult: of more than 800 recommendations assessed across four countries, only 17 could be reliably measured using hospital administrative data.²⁴

On the flip side, underuse is measured through “necessary care indicators” and “avoidable outcome indicators.” The first type checks whether patients with a given condition received recommended care, like annual eye exams for people with diabetes. The second type tracks diagnoses that should occur less often when patients have adequate access to care, essentially using bad outcomes as a signal that needed care was missed.²⁵ Globally, measurement of unmet health care need in older adults uses survey questions about whether people wanted care but could not get it, though question wording and response categories vary widely across studies.²⁶

Risk Adjustment and Fair Comparisons

Raw utilization numbers are nearly useless for comparing hospitals, regions, or health plans because the populations they serve differ. A hospital that treats older, sicker patients will naturally have higher utilization. Risk adjustment attempts to level the playing field by accounting for patient characteristics. Methods vary: some use mortality risk, others use diagnosis codes or prescription data to estimate expected illness burden, and still others rely on self-reported health status.²⁷

The choice of risk-adjustment method can change the rankings of facilities. A study comparing pharmacy-based and diagnosis-based adjustment models in VA hospitals found that the two approaches ranked the same eight facilities quite differently. There was more consistency among diagnosis-based models than between diagnosis-based and pharmacy-based ones.²⁸ This means that a hospital might look like a high performer under one adjustment system and a middling one under another. Anyone reading a utilization report card should pay attention to which risk-adjustment method was used.

Social Determinants and Utilization Patterns

In recent years, researchers have increasingly looked at how social circumstances, not just insurance and income, shape utilization. Housing instability, life dissatisfaction, stress, and transportation barriers have all been linked to higher use of health services, particularly emergency and inpatient care.²⁹ A large academic medical center study found that patients with housing instability or homelessness had roughly 34-49% more encounters with the health system than housed patients, and those in the most disadvantaged neighborhoods had 55% more encounters than those in the most advantaged areas.³⁰

More encounters does not mean better health. Often it means the opposite: people facing housing instability or other social challenges tend to cycle through emergency departments and hospitals because they lack stable access to primary and preventive care. This is one reason that simple utilization counts, taken alone, are a poor proxy for the quality of care a population receives.

Telehealth and Newer Measurement Challenges

Telehealth has introduced a new category of encounter that traditional measurement tools were not built for. Before the pandemic, surveys tried to capture telehealth use through questions about video and telephone visits, including whether the provider was familiar or new.³¹ The rapid expansion of telehealth during COVID-19 forced billing systems to create new codes and modifiers, which improved claims-based tracking but also created messy transitional data.

There is evidence that telehealth does not simply add visits on top of existing utilization. One analysis found that a telehealth visit reduced subsequent outpatient visits by about 14% in the following 30 days.³² That suggests some telehealth encounters replace rather than supplement in-person care, but the substitution pattern varies by condition, patient population, and how well the telehealth visit resolves the problem.

Post-Acute Care and the Gaps in the Data

What happens after a hospital discharge is increasingly recognized as a critical piece of the utilization picture. Post-acute care includes skilled nursing facilities, inpatient rehabilitation, and home health visits. Medicare Advantage enrollees tend to receive less post-acute care of all types compared to those in traditional Medicare.³³ Whether that reflects more efficient care management or inadequate service provision depends on patient outcomes, which brings us back to the need for utilization data linked to health results.

Frailty complicates post-acute measurement in older adults. Among patients discharged from skilled nursing facilities, those with moderate to severe frailty had about 71% higher odds of spending less time at home in the following six months compared to non-frail patients, even after accounting for age, sex, cognitive function, and other factors.³⁴ “Time at home” is itself an emerging utilization metric, reflecting the idea that what patients care about is not just surviving a hospitalization but returning to and staying in their normal life. It flips the traditional measurement perspective: instead of counting how much of the health system a patient consumed, it measures how much of their life they got to keep.

Barriers to Mental Health Service Use

Mental health utilization has its own measurement difficulties on top of the ones shared with medical care. People underreport behavioral health visits more than they underreport physical health visits, partly because of stigma and partly because the boundary between formal and informal mental health support is blurry. A large national survey in Bangladesh found that nearly three-quarters of respondents reported at least one barrier to mental health service use, with social concerns and perceived stigma being the most common obstacles.³⁵ People with anxiety disorders were about 75% more likely to report barriers than those without, which creates a paradox: the people who most need services face the steepest obstacles to using them.

This means that utilization statistics for mental health care almost certainly understate true need. Any measurement effort that relies solely on counting completed visits is systematically missing the people who wanted care but could not get past the barriers. Researchers are increasingly supplementing visit counts with surveys that explicitly measure unmet need, asking not just “did you see a provider?” but “did you want to see one and fail to?”