Table 1

Study variables and definitions

Data source

Conceptual definition

Operational definition


Dependent variables

Fall rate

Incident reports/fall incident report database

The fall rate was defined as the rate at which patients fall during hospital stays/1000 patient-days [15]. A fall was defined as an unplanned descent to the floor with or without injury. All falls are included, whether they result from physiologic or environmental causes [28].

(counts of total falls × 1000)/(total patient-days)

Injurious fall rate

Incident reports/fall incident report database

The injurious fall rate was defined as the fall rate/1000 inpatient-days at which physical injury occurs, regardless of severity (including minor/resulted in application of a dressing, moderate, major injury, and death) [15,28].

(counts of injury falls × 1000)/(total patient-days)

Covariates

Hospital

As identified by each study hospital

Four hospitals served as study sites: one academic medical center (Hospital 1); 2 teaching hospitals (Hospitals 2 and 4); and one community hospital (Hospital 3). This study used 3 dummy variables to capture 4 study hospitals, instead of hospital characteristics (e.g., bed size and teaching status), to control for the variations across study hospitals.

3 dummy variables were included in the regression model. Hospital 1 was used as a reference group; Hospital 2: 1 = Hospital 2, 0 = all other hospitals; Hospital 3: 1 = Hospital 3, 0 = all other hospitals; Hospital 4: 1 = Hospital 4, 0 = all other hospitals. When the values of Hospital 2, Hospital 3, and Hospital 4 = 0, Hospital 1 would be identified.

Unit type

As identified by each study hospital

The unit classification of each study unit, as designated by the hospital, was identified by the designated site coordinate. The unit type included 3 categories: (1) medical, (2) surgical, and (3) medical-surgical combined.

Two dummy variables were included in the regression model. Medical units were used as the reference group (unit type 2: 1 = surgical unit, 0 = all other units; and unit type 3: 1 = medical-surgical combined unit, 0 = all other units). When the values of unit type 2 and unit type 3 = 0, medical units would be identified.

Total nursing hours per patient-day (HPPDs)

The payroll database

As a system-centered measure, this was defined as the number of productive hours worked by nursing staff with direct care responsibilities per patient-day [15].

Total nursing hours/total patient-days

Percentage of the total nursing HPPDs supplied by registered nurses

The payroll database

As a system-centered measure, this was defined as the percentage of the productive nursing HPPDs worked by RNs with direct care responsibilities to the number of total productive nursing HPPDs worked by nursing staff with direct care responsibilities [15].

(Total nursing HPPDs supplied by RNs/total nursing HPPDs) × 100%

Patient age in percentage of patients aged ≥ 65 years

Patient management database

The percentage of all patients discharged from the study unit during the defined time period, who were ≥ 65 years.

(Sum of the years of the discharged patients in age/total discharged patients) × 100%

Case mix index (CMI)

Patient management database

CMI value mean of all patients discharged from the study unit during the defined time period represents the average diagnosis-related group relative weight for that unit-month. The CMI value is used to define the average acuity for patients admitted to a particular hospital [26].

(Sum of the CMI values of the discharged patients/total discharged patients) × 100%

Percentage of patients with altered mental status

Chart review of the nursing notes at admission

The percentage of patients hospitalized at the study unit on the 15th of the first month of each quarter, who had cognitive impairment or altered mental status. The charts of 10 randomly sampled patients per study unit were reviewed. If any cognitive impairment or altered mental status was identified in the chart at admission, this patient was coded as Yes (1); otherwise, No (0) was coded.

(Number of patients with cognitive impairment or altered mental status/10) × 100%

Percentage of patients with hearing problem

Chart review of the nursing notes at admission

The percentage of patients hospitalized at the study unit on the 15th of the first month of each quarter, who had hearing problems. The charts of a total of 10 randomly sampled patients per study unit were reviewed. If any hearing problems (with or without correction) were identified in the chart at admission, this patient was coded as Yes (1); otherwise, No (0) was coded.

(Number of patients with hearing problems/10) × 100%

Call light use rate per patient-day

The reports generated from the call light tracking system adopted by each hospital

Patient/family-initiated calls made from the pillow speaker or call cord attached to the wall in the patient rooms are categorized as normal calls (excluding the calls initiated in the bathrooms). In this project, only normal calls were studied. The normal call count will include all the calls either cancelled at the console or at the stations of origin (i.e., the patient's room).

(Counts of normal calls/number of the covered days) × (total number of days for the mo.)/(total patient-days for the month)

Due to the skewed distribution, this continuous variable was recorded into 10 equal groups and labeled in percentiles (10 = least frequent, 100 = most frequent). The recoded variable was analyzed as a continuous variable.

Predictor

Average response time to call lights

The reports generated from the call light tracking system adopted by each hospital

Patient/family-initiated calls made from the pillow speaker or call cord attached to the wall in the patient rooms are categorized as normal calls (excluding the calls initiated in the bathrooms). In this project, only normal calls were studied. The response time was defined as the time elapsed from normal call activation to call cancellation from the patient room. The response times were aggregated at the unit level for each month, and calculated by: (call light response time in seconds for all the calls made for the unit and month)/(total number of calls for the unit and month).

The average time for "Staff Response" on the reports generated from the call light tracking system was calculated as: (Sum of the call light response time for the calls in seconds)/(total call light use).

Due to the skewed distribution, this continuous variable was recorded into 10 equal groups and labeled in percentiles (10 = fastest, 100 = slowest). The recoded variable was analyzed as a continuous variable.


Tzeng et al. BMC Health Services Research 2012 12:84   doi:10.1186/1472-6963-12-84

Open Data