Data and Descriptive Overview

law by both nursing homes and patients. For example, nursing homes might substitute away from other inputs into quality of care as they increase their nurse employment and patient sorting to nursing homes might change as well. To the extent that patient sorting by unobserved acuity changes in such a way that more ill patients become more likely to sort into nursing homes with low initial staffi ng levels and thus bigger increases in staffi ng after the law becomes effective, this might lead to no apparent effect of staffi ng on patient outcomes. In the results section below, I show that there appears to be no effect of the law on observable measures of either facility- level fac- tor substitution or patient level sorting. While the potential for unobserved responses exists, the scant prior literature on this suggests the magnitudes of such a response are likely to be very small. While it might seem natural to use this setup to study the causal effect of total staff- ing all nurse hours on patient outcomes through an instrumental variables design— for example instrumenting the change in total nurse staffi ng with GAP i —I eschew such an approach here. In this context it is very likely that the causal effect of a change in staffi ng will depend heavily on the mechanism that induces such a change, violating the so- called “stable unit treatment- value assumption” SUTVA discussed by Rubin 1986. Put simply, one can easily imagine that a different minimum staffi ng law that required increases in higher- skilled nurses RNs would have quite different effects on patient outcomes than the California legislation. While the California legislation is similar to legislation adopted or proposed in many other states and so the results here have broad applicability, it may be inappropriate to use the results to infer the causal effects of different types of staffi ng increases. Keeping these concerns in mind, the two- stage least squares estimates can be easily inferred from the “fi rst- stage” esti- mates of the effects on nurse staffi ng and the “reduced form” estimates of the effect of the law on patient outcomes presented below.

IV. Data and Descriptive Overview

This paper uses data from two main data sources. Data on staffi ng and facility characteristics are taken from administrative data used to certify and license nursing homes for Medicare and Medicaid eligibility collected by the California Of- fi ce for Statewide Health Planning and Development OSHPD. The most important elements used from this data are the staffi ng levels at each facility, reported as the total number of hours worked by occupation over the relevant reporting period. For nurses, staffi ng levels are reported separately for supervisors and management, geriatric nurse practitioners GNPs, registered nurses, licensed vocational nurses, and nurse aides. I compute the facility’s total number of nursing staff hours as the sum of all hours worked by nonmanagement RNs, GNPs, LVNs, and NAs. 12 I divide this measure by the reported number of patient days to compute the facility’s HPRD staffi ng level. These data are then matched to quality of care measures from the federal Online Survey Certifi cation and Reporting System OSCAR data, collected by the Centers for Medicare and Medicaid Services CMS. These data are used to determine whether 12. In practice, due to the small number of hours involved the estimates are little affected by including these hours. facilities are complying with federal regulations, and facilities are surveyed for com- pliance at least every 15 months, though surveys are typically conducted at 12- month intervals. Harrington et al. 2000 argue that these data are accurate and reliable mea- sures of nursing home quality, and an IOM panel on nursing home quality recom- mended more intensive use of OSCAR data in future research IOM 2001. In the analyses presented here, I focus on two important patient outcomes in the OSCAR data where some adjustment for patient acuity at admission is feasible: the fraction of residents with pressure sores excluding Stage 1 that were not present on admission, and the fraction of residents with contractures that were not present on admission. Pressure sores also known as bed sores or decubitus ulcers are injuries to skin tissue caused by constant pressure caused, for example, by lying or sitting station- ary in the same position. Stage 2 or higher sores involve open wounds that risk infec- tions, and can be life- threatening if not properly treated. Contractures are shortenings of muscles or tendons, usually resulting from a lack of use of a joint and leading to loss of motion in that joint. Both of these measures are frequently used in measuring quality of care in nursing homes, because they are thought to be preventable if patients with mobility problems are aided in moving about or preventive exercises, care that is usually provided by nurse aides. I also examine the effects of minimum staffi ng regulations on two indicators for modes of care: the fraction of residents physically restrained who were not admitted with orders for restraints, and the fraction of residents with in- dwelling or external catheters not present on admission. These types of care are often considered substan- dard, as they generally are used as substitutes for nursing care such as help with toilet- ing Zinn 1993. Indeed, Cawley, Grabowski, and Hirth 2006 fi nd that nationwide, nursing homes in high- wage markets tend to use more “materials- intensive” modes of care use of catheters rather than nurse aides to assist incontinent patients. Similarly, higher nurse- staffi ng ratios might induce less “materials- intensive” care. Finally, I use OSCAR data on defi ciency citations as an overall measure of quality of care in the facility. Defi ciencies are given by surveyors on regularly scheduled visits for failure to comply with any of some 179 specifi c standards of care. I focus on total defi ciencies, and the subcategory “quality of care” defi ciencies that includes requirements to pre- vent pressure sores, falls, and physical decline. The total fi gure includes these defi cien- cies, and those related to other requirements covering quality of life resident rights issues, and administrative and record keeping practices Harrington et al. 2000. To generate an analysis sample, I begin with the universe of private skilled nursing facilities including investor and church owned submitting data to OSHPD in each year from 1994 to 1998. I start with these 1,133 fi rms to avoid analyzing new entrants, whose staffi ng ratios often fl uctuate wildly in their fi rst few years of operation as utilization ramps up. 13 I drop six facilities with extremely high staffi ng levels, and attempt to match the remaining 1,127 fi rms to the OSCAR data. 125 of these fi rms did not have adequate data on preperiod outcomes so were dropped, leaving 1,002. To simplify the analysis further, I restricted the analysis sample to the 965 fi rms re- maining in the sample in each year through 2004. I verify in Appendix Table A2 that attrition is not related to staffi ng levels in 1997–98, so attrition should not bias the 13. Of 1,166 fi rms present in the data in 1998, 1,133 of them were present in each year between 1994 and 1998 so this restriction excludes very few facilities. results below. The Appendix describes variable defi nitions, matching procedures, and analyzes sample selection in greater detail. Skilled nursing facilities, or nursing homes, provide both long- term medical and custodial care to their residents. Nationwide, residents are predominantly elderly—of approximately 1.5 million residents nationwide in 2004, 88.3 percent were aged 65 years or older—though younger individuals may spend some time recovering from surgery or a major accident in a nursing home. 14 Stays in nursing homes are gener- ally long: The median number of days since admission for 2004 residents was 463 days, with a mean of 835 days. More than half of residents require assistance with all fi ve “activities of daily living” ADL—bathing, dressing, toileting, transferring, and eating. Table 1 shows some descriptive statistics for the 965 nursing homes in California used in this analysis. The fi rst column shows the overall mean of various variables, and the next four columns show means for fi rms in each of four “quartiles” of the average HPRD staffi ng distribution for 1997 and 1998. Facilities had 101.4 beds on average, ranging from a low of 19 beds to a high of 391, with only slight differences across staffi ng quartiles. Total healthcare expenditures averaged about 4 million in 2005 dollars, and facilities employed an average of 61 people, with an obvious gradi- ent across staffi ng level. The overall production process is quite labor intensive, with nurse salaries and benefi ts accounting for about 40 percent of all expenditures. Care in nursing homes is provided nearly entirely by nurses, though doctors may supervise the development of treatment plans and patients may be transferred to a hospital emergency department in life- threatening situations. Most care—two- thirds of the total hours worked by nurses—is provided by NAs in the form of assisting residents with the activities of daily living listed above. LVNs and RNs account for 18 and 10 percent of all nurse hours worked, respectively, and are staffed primarily to develop and supervise medical treatment protocols. There is a clear skill gradient across these nursing occupations that is refl ected both in the amount of training necessary to get the corresponding occupational license, and in the average wages of each occupation. RNs are typically required to complete be- tween two and six years of postsecondary education, whereas LVNs typically require only one year. Nursing aides require no formal training to be hired, but must complete 100 hours of on the job training, and 50 hours of classroom training to be certifi ed, and must pass a state medical exam within four months of being hired. 15 In 1999, RNs made an average hourly wage of 23.84, followed by LVNs at 18.29 and NAs at 9.50. In some ways, the nursing home sector is an ideal place to study the effect of higher nurse- patient ratios on patient outcomes. The fact that nearly all care is administered by nurses means that for many patient outcomes, the quantity and quality of nursing labor may be the most important input in production. This contrasts with the hospital setting, where patient outcomes are likely also heavily infl uenced by the quality of doctor and equipment inputs. In terms of measurement, there may also be better cor- respondence between data on staffi ng and effective staffi ng for patients, since there is less segregation of patients and staff across subunits of a nursing home. 14. Statistics in this paragraph are taken from Jones et al. 2009. 15. This information is taken from California Employment Development Department LaborMarketInfo web- site: http: www .labormarketinfo.edd .ca .gov . T he J ourna l of H um an Re sourc es 42 Table 1 1997–98 Descriptive Statistics for LTC Facilities in Analysis Sample by Staffi ng Level Quartiles of 1997–98 HPRD Distribution Facility Descriptives All 1st 2nd 3rd 4th Number of beds 101.4 91.3 101.7 108.2 103.9 [49.3] [41.2] [46.3] [49.6] [56.3] Total healthcare expenditures 1,000s 3,925.70 3,043.30 3,646.00 4,110.70 4,755.60 [2053.4] [1,615.5] [1,673.4] [1,863.9] [2,453.9] Number of total employees 100.5 78.3 94.8 105.5 120.0 [47.4] [32.0] [40.6] [42.2] [58.0] Number of direct care nurses 61.0 47.9 58.0 65.7 70.7 Percent of total nursing hours by RNs 11.0 9.1 11.0 11.3 12.3 Percent of total nursing hours by LVNs 17.4 19.3 17.2 16.8 16.4 M at suda ira 43 Utilization Percent occupancy 88.4 89.1 88.5 88.6 87.4 Percent patient days paid by MediCal 67.7 78.8 72.1 67.0 52.8 Percent patient days self- paid 23.5 11.7 19.0 21.4 39.4 Resident outcomes Percent with pressure sores not present on admission 3.9 3.3 3.8 4.0 4.3 Percent with contractures not present on admission 7.3 5.6 6.8 8.8 8.0 Percent with catheters not present on admission 1.7 1.4 1.5 2.1 1.6 Percent with restraints not present on admission 14.6 13.6 14.5 15.6 14.6 Total defi ciencies regular surveys only 11.4 11.7 11.3 12.0 10.8 [6.4] [5.1] [6.0] [7.2] [6.9] Quality of care defi ciencies 2.9 3.0 2.8 3.1 2.7 [2.0] [1.8] [1.7] [2.2] [2.1] Number of facilities 965 230 232 232 271 Notes: The statistics shown are the averages across fi rms of the fi rm- level average value of each variable in the 1997 and 1998 period, ignoring missing data. Standard deviations of the fi rm- level average of some variables are in brackets. There are 965 fi rms overall divided into “quartiles” with respect to their average nurse HPRD in the 1997 to 1998 period. The fi rst three quartiles contain the 230, 232, and 232 fi rms, respectively, with the lowest staffi ng levels. All 271 fi rms in the 4th quartile had staffi ng levels already in compliance with the 3.2 HPRD threshold taking effect in 2000, whereas fi rms in the fi rst three quartiles all had staffi ng levels below the threshold.

V. Results