Making difficult decisions: the role of quality of care in choosing a nursing home.

Objectives. We investigated how quality of care affects choosing a nursing home.

Methods. We examined nursing home choice in California, Ohio, New York, and Texas in 2001, a period before the federal Nursing Home Compare report card was published. Thus, consumers were less able to observe clinical quality or clinical quality was masked. We modeled nursing home choice by estimating a conditional multinomial logit model.

Results. In all states, consumers were more likely to choose nursing homes of high hotel services quality but not clinical care quality. Nursing home choice was also significantly associated with shorter distance from prior residence, not-for-profit status, and larger facility size.

Conclusions. In the absence of quality report cards, consumers choose a nursing home on the basis of the quality dimensions that are easy for them to observe, evaluate, and apply to their situation. Future research should focus on identifying the quality information that offers the most value added to consumers. (Am J Public Health. Published online ahead of print March 14, 2013: e1-e7.)

Quality of care in nursing homes continues to concern patients, their families, and policymakers.[1] The attributes of quality that are important to long-stay nursing home residents and their families are often different from and more diverse than those that are important to hospitalized persons and their families in acute medical care settings.[2] Because it is a home and not a temporary residence, the physical environment of the nursing home plays an important role in resident and family judgments regarding such things as safety, privacy, freedom, and mobility, which affect general quality of life. Appropriate clinical care can help prevent and treat pressure ulcers, malnutrition, incontinence, declines in activities of daily living, pain, and falls.[3] Thus, the concern about nursing home quality encompasses both clinical care services and the hotel aspects of the social and physical environment.

In 2002, the Centers for Medicare and Medicaid Services (CMS) initiated the publication of the federal nursing home report card, Nursing Home Compare.[4] This report card provides information about several clinical dimensions of quality, such as change in activities of daily living, prevalence of pressure ulcers, and use of physical restraints, and allows consumers to compare rankings of these measures for each facility and with regard to national and state averages. Studies that examined the impact of the nursing home report card show quality improvements at the nursing home level in response to the publication and even a shift of resources from hotel services to support clinical dimensions of quality.[5-8] However, it is not clear that the publication of the nursing home report card has had any effect on consumers' choice of a nursing home.

Although some studies show that high clinical quality increases a nursing home's market share for short-stay residents,[9] others show that the nursing home report card has not affected consumer demand.[10] Although nursing home consumers may care as much about the clinical as the hotel aspects of care, their ability to assess them is not the same. Evaluation of hotel services often is easier and less costly; it can be accomplished by personal visits to the facility or by friends' or acquaintances' prior experience and reports. However, evaluation of clinical care quality is typically more difficult because it may require specialized knowledge and broader opportunities to observe the facility and access information about it.

Many consumers will not understand the relationship between processes of care and care outcomes, be able to factor in the impact of their individual risk profile, and anticipate what their clinical care needs will be when they are in the nursing home.[11] Thus, they may find it very difficult, if not impossible, to assess nursing home clinical care quality on their own -- or even to evaluate it using the current federal quality report card. In fact, studies find that these resources are not being widely used when selecting a nursing home.[12][13]

We examined quality of care as a factor in choosing a nursing home from among its competitors once the decision to enter a nursing home was made. We focused on 2 dimensions of quality of care to investigate the impact of their differential observability on choice. We have used the term "clinical care" to refer to all aspects of personal care and clinical and therapeutic services and "hotel services" to refer to room and board and aspects of care that relate to the physical and social environment and that affect safety and quality of life. We hypothesized that consumers' choice of nursing homes would be more sensitive to the aspects of quality that can be more easily evaluated or observed -- that is, hotel services -- than to the masked and more difficult to interpret clinical dimensions of quality.

Our study extends the literature in 2 ways. First, most previous studies have focused on the choice of nursing home versus alternative care settings[14-17] rather than the choice of a nursing home from among its competitors. The few studies that have examined choosing nursing homes for long-term care relied on surveys or small samples[18-20] or were not consistent in their findings.[9][10] We analyzed observed choice patterns of all long-term nursing home residents in 4 large states, thus offering more robust and generalizable information. Second, to date, none of the studies that examined the impact of quality on choice differentiated between the clinical and hotel dimensions of quality. We hypothesized that different dimensions of quality would affect choice differently because of the differences in consumers' ability to access, understand, and evaluate them. Examining this issue empirically offers new insights that are not available from prior studies.

We examined the choice of the consumer -- the prospective nursing home residents or their agents (e.g., family members or health care proxies) -- of a particular nursing home from an available set of facilities. We assumed that consumers choose the nursing home that maximizes their well-being from among all nursing homes in their choice set. We assumed that residents' well-being depended on the characteristics of the nursing home, that is, observable and masked quality dimensions; distance from prior residence, which could affect relationships with persons in or near that location; and size of the nursing home, as larger nursing homes are more likely to provide a wider range of services.

We also assumed that consumers' well-being depends on other nursing home facility characteristics, such as ownership status, occupancy rate, membership in a chain, and hospital affiliation, as these characteristics might indirectly serve as signals of nursing home quality that are difficult to observe directly. For example, nursing homes with nonprofit status may be viewed as higher quality and, therefore, more desirable to consumers.

We chose to study nursing home choice decisions in 2001, the last full year before the publication of the federal nursing home report card, Nursing Home Compare. At this time both hotel services and clinical care quality dimensions were not publically reported and consumers had to evaluate them directly. Furthermore, in 2001 there were no local or state-specific report cards available in the states participating in this study.

We also note that whereas there were no publicly reported clinical quality measures in 2001, we, as the analysts, did have access to data that were used in later years to calculate the clinical quality measures and thus could simulate the information that the CMS report card eventually made available to consumers. Thus, we were able to test nursing home choice with respect to the 2 types of quality in a period when consumers faced substantial differences in their ease of assessment.

We used 3 data sources. Data from the Minimum Data Set -- which includes assessments of individual nursing home residents on admission and every 3 months afterward as well as at times of significant change in health status -- contain individual-level information on all Medicare and Medicaid certified nursing homes. We used the Minimum Data Set to calculate the clinical care quality indicators as defined by CMS and to obtain individual-level characteristics.

The Online Survey Certification and Reporting data include nursing home operational characteristics and information on deficiency citations. State regulators issue deficiency citations during an annual survey of the facility if they determine that the facility does not meet specific regulatory quality standards. We used Online Survey Certification and Reporting data to obtain facility characteristics and to calculate hotel services quality measures.

We used the Medicare Denominator File to obtain the zip code of residence for new nursing home residents in the year before nursing home admission to calculate the distance between prior residence and the facilities in an individual's choice set.

We limited the study sample to private pay residents, as we expected the behavior of private pay and Medicaid residents to differ because of Medicaid residents' more limited choices.[21-23] We also excluded new admissions with a prior nursing home stay within 30 days or less. This eliminated choices by residents and families who may have had firsthand experience and information about the masked dimensions of nursing home quality, gained during a previous stay. We chose a 30-day window for prior admission because in the data available to us 75% of all new admissions with a prior nursing home stay had fewer than 30 days between the 2 stays.

Finally, the study population included all new long-term care nursing home admissions of private pay residents aged 65 years and older in New York, California, Ohio, and Texas during calendar year 2001. We chose these 4 states because they were among the 7 states with the largest number of nursing homes in the country and they included more than one quarter of all facilities in the United States.[24] Furthermore, these states varied in their regulations and level of quality,[25][26] thus affording higher generalizability.

Nursing home characteristics include hotel services and clinical care quality dimensions. To create the clinical quality indicators, we used the CMS algorithm and applied it to 2001 Minimum Data Set data.[27]

We measured clinical care quality using 3 quality measures developed by CMS and used in the original nursing home report card: percentage of residents with (1) decline in activities of daily living since admission (calculated by dividing the number of residents with a lower activities of daily living score in their first quarterly assessment compared with the admission assessment by total number of eligible residents), (2) pressure sores (calculated by dividing the number of residents with pressure sores during the assessment by total number of eligible residents), and (3) physical restraints (calculated by dividing the number of residents who were physically restrained daily during the time of assessment by total number of eligible residents).

CMS's original report cards included 2 additional quality measures: percentage of residents with infections and percentage of residents with pain. We did not include these 2 measures in our study, as the infections measure was not stable and was dropped later from CMS's report card, and the pain measure using CMS's algorithm did not yield the same results as those originally posted on CMS's Web site.

We measured hotel quality, the second quality dimension, using Online Survey Certification and Reporting data and included 42 quality deficiencies that relate to hotel services, the domain we considered potentially observable by consumers during a nursing home walk-through or during an informational meeting with the nursing home staff. We used deficiency citations rather than nursing home's expenditures on hotel services, as they include measures that cannot always be bought, such as staff friendliness and providing care that supports residents' dignity and self-respect (Table 1 provides a detailed list).

We counted each of the 42 measures only if we found a deficiency related to that measure. Similar to other studies, which used all deficiencies, we defined hotel services quality as the unweighted sum of these deficiencies.[28] To normalize for the size of the nursing facility, we divided the sum of deficiencies by the number of beds in the facility, because larger facilities have a higher exposure to citations (e.g., 10 citations in a 20-bed facility is different from 10 citations in a 200-bed facility).

Lacking information about clinical quality of care, consumers may interpret not-for-profit ownership status as a signal of quality.[29] Thus, we included ownership status in the multivariate model (for-profit, public, and not-for-profit as the reference group). Previous studies show that nursing home residents prefer to stay close to their community, friends, and family members.[16][18][20][30] Therefore, we included a distance variable, defined as the straight-line distance from the centroid of the zip code of prior residence to the centroid of the zip code of each nursing home in the individual's choice set. We included additional facility characteristics that may also affect the choice decision: facility size (average number of licensed beds), occupancy rate, chain membership, and hospital ownership.

To identify the relevant choice set for each nursing home admission, we analyzed new nursing home admissions over a 3-year period, 2001-2003, using patient flows, following Zwanziger et al.[31] The analysis included 2 steps. In the first step, we determined the core market for each nursing home on the basis of the zip codes contributing the first 70% of admissions to that nursing home. In the second step, we determined the choice set for each patient by identifying all the nursing homes that compete in the patient's zip code of residence. In other words, the choice set for each patient included all the nursing homes that had that patient's zip code in his or her core market.

Tag Description 0174 Have a private telephone available for use. 0175 Allow married couples to share a room. 0240 Provide care for each resident in a way that keeps or builds the resident's quality of life. 0241 Provide care in a way that keeps or builds each resident's dignity and self-respect. 0242 Make sure each resident has the right to choose activities, schedules, and health care according to his or her interests, assessment, and plan of care. 0243 Make sure that private space is available in the nursing home for resident groups or residents' families to meet. 0244 Listen to the resident or family groups and act on their complaints or suggestions. 0245 Make sure that each resident has the right to join in social, religious, and community activities. 0246 Provide services to meet the needs and preferences of each resident. 0248 Provide activities to meet the needs of each resident. 0249 Hire a qualified activities director. 0250 Provide social services for related medical problems to help each resident achieve the highest possible quality of life. 0251 Hire a qualified social worker. 0252 Keep safe, clean, and homelike surroundings. 0253 Provide needed housekeeping and maintenance. 0254 Provide clean bed and bath linens that are in good condition. 0255 Provide closet space in each resident's room. 0256 Keep adequate and comfortable lighting in all areas. 0257 Keep temperature levels comfortable and safe. 0258 Keep sound levels comfortable. 0323 Make sure that the nursing home area is free of dangers that cause accidents. 0360 Provide a tasty and well-balanced diet that meets the nutritional needs of each resident. 0363 Make sure that residents are well nourished. 0364 Prepare food that is nutritious, appetizing, tasty, attractive, well cooked, and at the right temperature. 0365 Provide food in a way that meets a resident's needs. 0366 Offer other nutritious food to each resident who will not eat the food served. 0368 Provide 3 meals at regular times, serve breakfast within 14 h after dinner, and offer a snack at bedtime each day. 0369 Provide special eating equipment and utensils for each resident who needs them. 0371 Store, cook, and serve food in a safe and clean way. 0372 Dispose of garbage properly. 0454 Be designed, built, equipped, or well kept to protect the health and safety of residents, workers, and the public. 0457 Provide bedrooms that hold no more than 4 residents per room. 0458 Provide rooms that are big enough for each resident. 0459 Provide bedrooms that open into an exit hallway. 0460 Provide bedrooms that do not allow residents to see each other when privacy is needed. 0461 Make sure each resident has at least 1 window to the outside in a room, a room at or above ground level, adequate bedding, furniture that meets the resident's needs, and adequate closet space. 0462 Put a bathroom in or next to each resident's room. 0463 Make sure that a working call system is available in each resident's room or bathroom and bathing area. 0464 Provide at least 1 room set aside to use as a resident dining room and for activities that is a good size, with good lighting, airflow, and furniture. 0465 Make sure that the nursing home area is safe, easy to use, clean, and comfortable. 0467 Have enough outside ventilation. 0468 Put firmly secured handrails on each side of hallways.

Note. OSCAR = Online Survey Certification and Reporting.

Defining the patient choice set on the basis of nursing home market definition that relies on resident-origin zip codes is more accurate than is defining nursing home markets on the basis of county boundaries, as most previous studies of nursing homes have done.[32][33]

Following McFadden,[34] we used a conditional multinomial logit to model the choice of a long-term nursing home admission conditional on nursing home characteristics (choice attributes). Because we had a directional hypothesis -- that higher quality would increase the probability of choice -- we used 1-tailed tests and determined statistical significance at the P< .05 level. In addition, because each quality variable is in different units, we calculated standardized odds ratios (ORs) to be able to compare the effect size across quality and all other variables.

Because occupancy rate might be considered endogenous with quality, we also estimated all models excluding occupancy rates as a sensitivity analysis. We found the results to be robust to the exclusion of this variable and therefore we have presented only the full model.

Table 2 provides descriptive statistics for all independent variables included in the multivariate analyses. Overall, there were 18 185 residents and 1350 nursing homes in California; 5680 residents and 1028 nursing homes in Ohio; 4182 residents and 702 nursing homes in New York; and 4857 residents and 1207 nursing homes in Texas. The clinical care quality ranged from 13.6% to 14.8% of residents with a decline in activities of daily living, from 8.5% to 10.4% of residents with pressure sores, and from 7.4% to 17.9% of residents with physical restraints across all states in the study. The average hotel services quality ranged from 1.4 to 3.4 deficiencies per bed. Most nursing homes in each state were for-profit, and the average distance between the chosen nursing home and the individual's prior residence was 16 to 26 miles across all states. The average age of the nursing home residents was about 83 years in all states. The vast majority of nursing home residents were White (87%-92%) and had a high school education or above (71%-82%).

Variable California, No., %, or Mean (SD) Ohio, No., %, or Mean (SD) New York, No., % or Mean (SD) Texas, No., %, or Mean (SD) Nursing home characteristics Sample size 1350 1028 702 1207 Clinical care quality: % of residents with challenges Decline in activities of daily living 14.5 (8.1) 14.8 (6.7) 13.6 (5.2) 16.1 (6.5) Pressure sores 10.4 (7.3) 9.3 (4.6) 9.3 (4.2) 8.5 (4.9) Physical restraints 17.9 (11.7) 9.9 (7.9) 7.4 (8.1) 17.5 (10.0) Hotel services quality: no. of deficiencies for more observable quality dimension per bed 3.4 (2.5) 1.7 (1.7) 1.4 (1.7) 2.3 (2.0) Ownership status For-profit 76.30 74.00 48.70 82.20 Not-for-profit 20.10 23.00 43.30 15.20 Public 3.60 3.17 8.10 2.60 Nursing home characteristics Distance from prior residence in miles 15.59 (18.6) 20.28 (22.4) 16.38 (20.3) 25.54 (28.1) Number of licensed beds 101.00 (89.8) 117.00 (83.0) 186.00 (130) 106 (46.7) Occupancy rate 82.20 (17.8) 77.70 (21.0) 93.40 (9.7) 71.10 (19.8) Affiliation Part of a chain 63.7 57.6 17.7 73.5 Hospital affiliated 14.0 5.8 11.4 6.7 Individual characteristics-private pay new admissions Sample size 18 185 5680 4183 4857 Age 83.1 (7.7) 83.3 (8.0) 82.0 (7.9) 83.9 (7.6) % White 86.7 92.3 89.8 89.8 % High school graduate or above 82.2 70.9 70.9 73.3

Table 3 presents the estimated conditional logit model with the coefficient estimates for the quality variables: hotel services and clinical care. Note that for all quality variables, lower scores denote higher quality and ORs smaller than 1 imply that lower quality leads to lower probability of a facility being chosen. As hypothesized, we found that hotel service quality is significantly associated with the probability of a nursing home being chosen.

A 1 SD decrease in hotel quality lowered the nursing home's odds of being selected by about 20% in California, 11% in Ohio, 17% in New York, and 12% in Texas. These findings were statistically significant at the P<.05 level in all 4 states. By contrast, in all but 1 case (physical restraints in New York) none of the clinical measures of quality was significantly associated with the probability of a nursing home being chosen. This suggests that clinical quality of care, which residents and their families may find more difficult to observe and evaluate, did not directly influence choice of nursing homes.

As expected, nursing home choice was significantly associated (P<.01) with distance and other facility attributes, including size (1 SD increase in number of beds in the nursing home increased the nursing home odds of being selected by 20%-50%) and occupancy rate. In fact, distance was the strongest predictor of nursing home choice, suggesting that residents prefer to enter a nursing home that is closer to their prior residence. Hospital affiliation and chain participation were important in some but not all states.

Variable California Ohio New York Texas Hotel services quality 0.7990[*] 0.8885[*] 0.8343[*] 0.8805[*] Clinical care quality measures Decline in activities of daily living 1.1249 1.0807 1.0839 1.0863 Pressure sores 1.3804 1.0236 1.2908 1.0217 Physical restraints 1.0230 1.0115 09293[*] 0.9724 Ownership status For-profit 0.6586[**] 0.8433[**] 0.9974 08424[**] Public 0.5132[**] 1.3155[**] 0.5655[**] 0.5793 Nursing home characteristics Size 1.2666[**] 1.2839[**] 1.5001[**] 1.1969[**] Distance from prior residence in miles 0.0036[**] 0.0005[**] 0.0118[**] 00061[**] Occupancy rate 1.1723[**] 1.4491[**] 1.0795[*] 0.8701[**] Affiliation Part of a chain 1.0462[**] 1.0448 0.9907 1.0657 Hospital affiliated 0.4243[**] 1.1643 0.9179 0.0837[**] Pseudo R2 0.2299 0.3070 0.1868 0.3288

[*]P <.05; [**]P <.01 in a 1-tailed test.

Of particular interest are the findings related to ownership variables, which we hypothesized to influence choice as signals of quality. Specifically, we expected nursing home residents and their families to view for-profit ownership as signals of lower quality. It was less clear how public ownership might be viewed, as public nursing homes have historically played a dual role of catering to the indigent (suggesting poorer quality) and providing the most complex care (suggesting higher quality). As shown in Table 3, for-profit status decreased nursing homes' odds of being chosen by 16% to 34%, except in New York State, where ownership did not appear to matter, probably because New York has a much more stringent regulatory system than do other states, which prohibits the entrance of proprietary chains into the market. Furthermore, New York State has a much larger percentage of nonprofit facilities and, perhaps as a result, has a for-profit sector that is more similar to its nonprofit counterpart. These findings are consistent with our hypothesis that ownership plays an important role as a signal for clinical care quality as do those aspects of quality that consumers can observe.

We found that the probability of a nursing home being chosen was significantly associated with its hotel services rather than its direct measures of clinical care. The impact of hotel services was such that a decrease of 1 SD was associated with an 11% to 20% decrease in the odds of consumers choosing a nursing home. We observed this in 4 large states (California, New York, Ohio, and Texas) from different regions of the country with different regulatory systems and different levels of quality, suggesting that this finding might be generalizable to other US markets.

The lack of association between direct measures of clinical care and nursing home choice did not imply that clinical quality is not an important factor for consumers in considering a nursing facility.[35] Rather, it is consistent with our hypothesis that consumers do not as easily observe this dimension of quality, and therefore they cannot readily incorporate it into their decision-making process.[36] This leads to consumer choices that are not fully informed.[37] In fact, we found that consumers used indirect measures of clinical care that they can easily observe, such as ownership status, in choosing a nursing home. Thus, our findings also support our hypothesis that consumers interpret nonprofit ownership status as a signal for high quality, as shown in previous studies.[29][38] The fact that nursing homes with higher occupancy rates have greater odds of being selected may also suggest that prospective nursing home residents use occupancy rate as an additional signal for clinical quality.

This study offers 2 important pieces of empirical evidence. First, it demonstrates that consumers faced with choosing a nursing home consider quality of care. Second, consumers are more likely to consider those dimensions of quality they can easily observe or evaluate. The debate about the need for and the impact of quality report cards on referral choices has been ongoing. For example, a survey of nursing home residents and their families found that less than one third of family members of nursing home residents used the Nursing Home Compare report card.[12] Moreover, the main use was to find the location of the nursing home rather than to get information about the quality of the facility. Another study, using focus groups of nursing home residents and their families, suggests that the quality information available on the Internet is not always presented in a way that is easy for consumers to understand and does not always include the information that is important to them.[13] In light of our findings, we would argue that the majority of consumers may not be able to obtain relevant easy to interpret information about clinical care and, hence, cannot incorporate that information in their choice, even after the publication of the nursing home report card.

Although several studies have already demonstrated the impact that quality report cards have on patients' choices (e.g., for health maintenance organizations[39] and physicians[40]), this study offers a complementary and important new perspective. It suggests that quality report cards for nursing homes can be valuable to consumers primarily in providing information on clinical quality of care but only if provided in a way that is directly relevant and easy for them to evaluate. Thus, report cards should focus not only on providing the clinical aspects of care but also on presenting them in a way that is relevant and easy to interpret.

We note several limitations. First, when calculating the distance between each nursing home alternative and the prior place of residence, we assumed that the residents' zip code before the nursing home admission indicates actual place of residence. Although determining actual place of residence is desirable, administrative data did not allow us to ascertain that information.

Another limitation was the potential generalizability of our findings. Although we believe that our focus on 4 large and diverse states increased the likelihood that our findings will hold for other areas of the country, it is possible that cultural, educational, and market-related factors may shape consumer behavior in ways that would lead to different choices and a different emphasis on quality dimensions. We included only 2 dimensions of quality, assuming that hotel services are more observable to consumers than is clinical care. In fact in each quality dimension, some aspects can be more observable to consumers than are others (e.g., for the hotel quality dimension, cleanliness of the nursing home is more observable than is quality of life of the residents). It is important to understand that those quality dimensions are not absolutely "observable" or "unobservable" but rather have different grades of observability on a continuous scale.

Our study offers empirical evidence suggesting that consumers value quality when selecting a nursing home. Regardless of the availability of a quality report card, consumers choose a nursing home on the basis of the quality dimensions that are easy for them to observe, evaluate, and apply to their own situation. Further research to understand the value that residents place on information relevant to their choice of a nursing home would help in designing report cards that would be most useful to them at the time of greatest need.

Reprints can be ordered at http://www.ajph.org by clicking the "Reprints" link.

This article was accepted January 10, 2013.

Dana Mukamel gratefully acknowledges support from the National Institute on Aging (grant AG027420).

Approval for this study was obtained from the University of Rochester's institutional review board.