healthmedicinet

Study population

Older adults aged 65 and over, determined as of March 31
^st
each year, from the Saskatchewan’s person registry system (PRS), were constituents
of the “covered population” in Saskatchewan. The study population includes those with
provincial health insurance coverage, living in Saskatchewan anytime during the period
from the fiscal year 1995/96 to 2004/05, and were hospitalized with an injury in any
year. A fiscal year covers the period from April 1st to March 31st. A detailed exploration
of hospital separation data for all older adults hospitalized with an injury (?=?39,867) and those with a fall injury hospitalization (?=?30,757) was conducted. Information was not available for members of the Royal Canadian
Mounted Police (RCMP), members of the Armed Forces and inmates of the Federal Penitentiary
and non-Saskatchewan residents, as they were not eligible for provincial health insurance
coverage.

Research design and procedures

This surveillance study used Saskatchewan’s administrative health services data on
injury hospitalizations among individuals aged 65 years and over living in Saskatchewan,
Canada at any time in the 10-year period from 1995/96 to 2004/05. The Saskatchewan
Ministry of Health extracted the injury hospitalization dataset and de-identified
before releasing it for this study. De-identified records of injury hospitalizations
linked with person registry information on sex, dates of birth, and death available
from the Saskatchewan Ministry of Health’s hospital separations data were utilized.
To capture all records of injury hospitalizations over the period, the specific diagnostic
codes (ICD-9 and ICD-10-CA) based on International Statistical Classification of Diseases and Related Health Problems (World Health Organization 2004]) were used. Specifically, external causes of injury codes defined the types of injuries
recorded in the hospital separation database (Saskatchewan Ministry of Health 2008]). All of the ICD codes for external cause of injury (ICD-9 codes 880-889 and ICD-10
CA codes W00 to W19), including those for unintentional falls (E880-E888 and W00-W19),
were used. The ICD codes for injuries due to medical misadventure, i.e., injuries due to adverse effects of medical procedure (E870-E876, E878-E79 and Y60-Y69, Y71-Y84, Y88.1) or therapeutic agents (E930-E949 or Y40-Y59 or Y70-Y88) were excluded. Information obtained from the hospital
separation dataset included admission and discharge dates, injury hospitalization
type (in-patient or day surgery), sex, date of birth, death flag in a given year,
date of death, hospital visit type (in-patient or day surgery), dead or alive on discharge,
length of stay, age, ICD-codes for falls, and most responsible diagnosis for the hospital
stay.

Variables of interest

The variables of interest for this analysis were type of injury, cause of injury,
age group, sex, area of residence within the province (rural, urban, and north), and
year of injury hospitalization. Specifically, age groups (65–74, 75–84, and 85+?years),
sex (male and female), and the area of residence at the time of injury (rural, urban,
north) were selected as categorical variables for analysis. Urban area of residence
included all of Saskatchewan’s 13 cities (Estevan, Humboldt, Melfort, Melville, Moose
Jaw, Lloydminister, North Battlefords, Prince Albert, Regina, Saskatoon, Swift Current,
Weyburn, and Yorkton) with a total population of 78,642 older adults (65+) in 2001,
the year for which the covered population was considered as a proxy for average population
for use as the denominator for average of all 10-year data. The northern area of residence
encompasses the northern health regions of the province (Keewatin Yatthé, and Mamawetan
Churchill River Regional Health Authorities, and Athabasca Health Authority), each
with their own unique demography and health issues (Saskatchewan Ministry of Health 2008]). For the purpose of this study, except for northern region identified as northern
area, all cities in Saskatchewan represented the urban area, while rural area of residence
included the rest of Saskatchewan comprising towns, villages, and farms.

Data analysis

Records of all individuals aged 65 and over hospitalized with an injury (?=?39,867) were extracted from the dataset and classified into a category of “fall
injury” (1?=?yes, 0?=?no). Descriptive analysis of fall injury hospitalizations (?=?30,757) by age group, sex, and area of residence was conducted. Crude fall injury
hospitalization rates were calculated as the number of injuries by fiscal year (both
sexes) divided by the corresponding covered population (aged 65 and over) in Saskatchewan.
The covered population for the year 2001 was selected as the common denominator because
it was roughly in the middle of the period of review in the present study.

Logistic regression analysis was applied to all injury hospitalizations (?=?39,867) with a fall (yes?=?1) or without a fall (no?=?0) in order to determine
if the frequency of fall injury hospitalizations was associated with the following
variables (risk factors): age group, sex, and area of residence at the time of hospitalization
as the covariate. A stepwise logistic regression analysis was conducted on all fall-related
injury hospitalizations. Stepwise logistic regression analysis was selected as it
conducts forward selection or backward elimination of the factors as required for
the regression model used. Likelihood ratios were used to determine the overall model
significance of final regression models. Odds ratios (OR) and 95 % confidence intervals
(CI) were presented for significant risk factors. Somer’s D goodness of fit test was
conducted to assess the strength of pairs within the model. Strong associations among
the variables were highlighted for the regression analysis. A p value ?0.05 was considered significant for all statistical analysis (Lang and Secic
2006]).

A survival analysis of person-level data by year was conducted, using SAS LifeTest
Procedure that produces life tables and Kaplan-Meier survival curves, to determine
the probable time to death (event) for those who experienced a fall injury hospitalization
for the first time. By selecting the first record of injury hospitalization for a
patient in each year, the duplicate records are removed for the calculating the person-level
fall injury hospitalization rates and conducting survival analysis. Survival analysis
estimates the conditional probability of survival over time from the discharge date
of fall injury hospitalization in the database. The analysis accommodates right-censored
data where the end-point event has not yet occurred or is unknown (Laurent 2002]). The percentage of older adults still alive at the end of each time interval (every
5 years after age 65) was used to estimate the probability that a typical older adult
will be alive at the end of any given period. When graphed, these estimates formed
distribution (Kaplan-Meier) curves of the probabilities of survival for the different
time periods. Life tables present the number of failed, censored, effective sample
size and hazard ratios (probability per time unit that a case that has survived to
the beginning of the respective interval will fail in that interval) for older adults
hospitalized with a fall injury (Lang and Secic 2006]). Only the first instance of fall injury hospitalization ever was included in each
analysis to avoid duplication of those who might have had more than one fall injury
hospitalization over a 10-year study period. The case definition for the survival
analysis was the first hospitalization for a fall injury over 10 years under review
(analysis reflects impact of only first fall injury; not the possible multiple fall
injuries in database for a given individual). Kaplan-Meier survival analysis curves
were presented by age, sex, and area of residence and overall (mean) survival for
each analysis. Log-rank and Wilcoxon tests of homogeneity were conducted to assess
differences among strata for each risk factor (Lang and Secic 2006]). All analyses were performed using SAS software, Version 8 (SAS Institute, Inc.
Cary, NC, USA).