Hospital Morbidity Database (HMDB)

Status:
Inactive
Frequency:
Annual
Record number:
3203

The Hospital Morbidity Database (HMDB) is a national database containing demographic, administrative and clinical data on inpatient hospitalizations in Canada.

Detailed information for 1996 - 1997

Data release - February 24, 1999 (Responsibility for the database was transferred to the Canadian Institute for Health Information (CIHI).)

Description

Note:
As of the 1994/95 data year, responsibility for the database was transferred to the Canadian Institute for Health Information (CIHI). Public enquiries about the Hospital Morbidity Database should be directed to CIHI at morbidity@cihi.ca.

Consult the Canadian Institute for Health Information at the following Website (www.cihi.ca) for documentation on the Discharge Abstract Database and the Hospital Morbidity Database.

The Hospital Morbidity Database (HMDB) is a national database containing demographic, administrative and clinical data on inpatient hospitalizations in Canada.

In the early 1960s the Royal Commission on Health Services (the Hall Commission of 1961) recommended that the then Dominion Bureau of Statistics (Statistics Canada) collect and publish national hospital morbidity statistics. In order to produce comparable statistics at the national level, Statistics Canada created the Hospital Morbidity program (starting with the 1960 data year). Through this program, Statistics Canada obtained hospital data from the provincial ministries of health and produced a standardized national file that contained data elements that were common to all provinces. (See the Data sources section for more information.)

Statistics Canada was responsible for the HMDB for data years 1960 to 1993/1994. As of the 1994/1995 data year, responsibility for the database was transferred to the Canadian Institute for Health Information (CIHI). Statistics Canada continues to be responsible for public dissemination of data for the data years 1993/1994 and earlier while CIHI is responsible for public dissemination of data for 1994/1995 and later. CIHI provides annual updates of the database to Statistics Canada where it is used for analysis and the development of the Hospital Person Oriented Information Database (HPOI).

Until the transfer of the HMDB to CIHI, Statistics Canada routinely published annual statistics from the database. The annual publications provided provincial and national statistics on the number and rates for separations by diagnosis grouping, procedure grouping, sex and age groups, total days stay, average length of stay and days per 100,000 population. The last publication was for the 1993/1994 data year. (See the Documentation section for more information.)

Reference period:
Data years 1960 to 1978: Calendar year; Data year 1979-1980 to present: Fiscal year

Subjects

  • Children and youth
  • Diseases and health conditions
  • Health
  • Health and well-being (youth)
  • Health care services

Data sources and methodology

Target population

The conceptual universe for the Hospital Morbidity Database (HMDB) includes inpatient hospitalizations in Canada that occur in general, convalescent, rehabilitation and chronic hospitals. The database excludes hospitalizations in psychiatric hospitals and hospitalizations in federal facilities (for example, military hospitals, prisons and Indian reserves.)

Information on each inpatient hospitalization is collected at the time of the patient's separation from hospital. A separation includes discharges, sign-outs, transfers to another facility and deaths. The date of separation determines the data reference year for that hospitalization.

A hospitalization refers to a hospital stay in which the patient was officially admitted as an inpatient. (This usually means that the patient stayed at least one night in hospital.) Day surgery cases, outpatient or clinic visits and emergency room visits are not included in the HMDB.

Each record in the database represents an inpatient hospitalization. Statistics from this database are, therefore, based on hospital stays rather than unique persons. For example, a person who is admitted and discharged five times during the same fiscal year will have five records in the database for that data year.

As of fiscal 1994/1995

As of fiscal 1994/1995, the population of reference consists of inpatient hospitalizations that occur in acute care facilities in Canada. Although the Hospital Morbidity Database still contains hospitalization records from some non-acute care facilities, the level of coverage has been significantly decreasing since fiscal 1996/1997.

Sampling

This survey is a census with a cross-sectional design.

Data sources

Responding to this survey is mandatory.

Data are extracted from administrative files.

The Hospital Morbidity Database is compiled from administrative records that originate in the reporting hospital. Upon a patient's separation from hospital, an admission/separation form is completed. This form is used to document demographic, administrative and clinical information pertaining to that particular stay in hospital.

Historically, these forms were designed and processed by provincial hospital insurance commissions. As a result, the type of information collected and the level of processing performed on this information was province-specific and not standardized at the national level. In the early 1960s the Royal Commission on Health Services (the Hall Commission of 1961) recommended that the then Dominion Bureau of Statistics (Statistics Canada) collect and publish national hospital morbidity statistics. In order to produce comparable statistics at the national level, Statistics Canada created the Hospital Morbidity program (starting with the 1960 data year). Through this program, Statistics Canada obtained hospital data from the various provincial ministries of health. These data were submitted to various edits to ensure data quality and standardization. A Canada format is used in the HMDB, incorporating as many common elements as are necessary and possible in order to produce national statistics. As of the 1968 data year, information on surgical procedures performed during the hospitalization were collected and published.

Over a period of years an increasing number of provinces opted to use the services of a centralized data processing system operated by the Hospital Medical Records Institute (HMRI). This centralization increased efficiency and standardization among the participating provinces. In the mid 1990s, the activities of HMRI were assumed by the newly created Canadian Institute for Health Information (CIHI). The name of the data holding at CIHI that now maintains this information is the Discharge Abstract Database (DAD).

As of fiscal 1994/1995, the compilation of the national Hospital Morbidity Database became the responsibility of CIHI. The main source of data for the HMDB became the DAD, with the remaining data supplied by provincial ministries of health that did not belong to the DAD. As of fiscal 2005/2006, however, the only jurisdiction not reporting to the DAD is Quebec. Agreements are in place to routinely include these records in the HMDB.

Please see the Documentation section for more information on:
a) common data elements collected prior to 1994/1995 and
b) a history of the diagnosis and procedure classifications used in the HMDB.

Error detection

Prior to fiscal 1994/1995:

The data files that Statistics Canada received from the provinces had already been edited by whatever data collection system the provinces were using. Before inclusion in the Hospital Morbidity Database (HMDB), however, all records were submitted to a set of validity and correlation edits at Statistics Canada.

The validity edits checked that the necessary data elements were present, consistent and not duplicated. The correlation edits (also called the medical edits) checked that the reported diagnoses and procedures were consistent with the reported age and sex. Missing items were imputed according to a module based on past experience with the data. Edit failures, with changes made, were printed out as turn-around documents to be reviewed, accepted or changed and then fed back into the system.

Published statistics from the HMDB were based on the tabulating diagnosis and the tabulating procedure. Hospitals could report several diagnoses or procedures for each hospitalization, but only one diagnosis and one procedure per record could be selected for statistical tabulation. Edit processes at Statistics Canada ensured that the most appropriate diagnosis or procedure was selected.

As of fiscal 1994/1995:

The Canadian Institute for Health Information (CIHI) has been responsible for the data quality of the Hospital Morbidity Database since fiscal 1994/1995.

CIHI creates the HMDB from records in the Discharge Abstract Database and records from those provincial ministries of health that do not participate in the DAD. Records entering the HMDB have previously undergone data quality checks from their original source. The DAD has a comprehensive edit and correction system and non-DAD records undergo data quality checks from the submitting ministries of health (for example, Manitoba prior to fiscal 2004/2005 and Quebec). Prior to CIHI's release of the annual HMDB file, CIHI routinely submits the file to Statistics Canada for review and approval.

Contact CIHI for more information on error detection of the HMDB.

Imputation

Prior to fiscal 1994/1995:

Missing items (for example, age and sex) were imputed according to a module based on past experience with the data. Edit failures, with changes made, were printed out as turn-around documents to be reviewed, accepted or changed and then fed back into the system.

As of 1994/1995:

Age and sex are no longer imputed. Invalid or blank values are flagged as an error.

Quality evaluation

Prior to fiscal 1994/1995:

Quality evaluation was based on the review of the edit failures, producing frequencies for certain data elements and performing trend analysis.

Frequently, the number and rate of separation or days will exhibit a marked change from one year to the next. Extreme changes usually do not reflect real changes in the incidence or prevalence of a condition but rather some administrative artifact in the collection and processing of data at the provincial level. Unfortunately, it was not possible to keep abreast of all the changes that occurred in the provinces. Nor was it possible, in the time available between the receipt of processed and edited data and its submission for publication to investigate all apparent anomalies.

Differences in the rates of discharge and days of stay among the provinces and territories may be attributable to the degree of availability of alternative types of care such as day surgery and outpatient clinics. Differences in rates between years may also be attributable to changes in health care policies.

As of fiscal 1994/1995:

CIHI applies quality control at various levels. For example, CIHI provides support services and education programs for personnel capturing data in hospitals. As well, it works with software vendors to continually improve data submission applications. During the production of both the DAD and HMDB, CIHI applies extensive edit and correction processes. CIHI also conducts special studies such as re-abstraction studies.

As of fiscal 2001/2002, CIHI produces annual data quality documentation for the HMDB. The documentation includes general data limitations in terms of accuracy, comparability, usability and relevance.

Public enquiries about the quality of the Hospital Morbidity Database should be directed to the Canadian Institute for Health Information.

Disclosure control

Statistics Canada is prohibited by law from releasing any data which would divulge information obtained under the Statistics Act that relates to any identifiable person, business or organization without the prior knowledge or the consent in writing of that person, business or organization. Various confidentiality rules are applied to all data that are released or published to prevent the publication or disclosure of any information deemed confidential. If necessary, data are suppressed to prevent direct or residual disclosure of identifiable data.

Prior to fiscal 1994/1995

Statistics Canada is responsible for the public release of data from the Hospital Morbidity Database for data years 1960 to 1993/1994. Only tabular data at the provincial and Canada levels were included in the annual publications and cells with small counts were not suppressed. Custom requests for historical data now undergo disclosure control. Cell counts of 3 or less are suppressed and in order to avoid residual disclosure, other cells are chosen for suppression. Alternatively, aggregation of data at a higher level may be done to avoid the need for suppression.

As of fiscal 1994/1995:

CIHI is responsible for public dissemination of data from the Hospital Morbidity Database. Any analytical articles published by Statistics Canada that uses data from the HMDB remains subject to disclosure control.

Data accuracy

Prior to data year 1994/1995:

Before inclusion in the HMDB, all records were submitted to validity and correlation edits at Statistics Canada. An annual average of 1% of records had an error detected during the validation and correlation edits.

Coverage was estimated at almost 100% for the provinces. The survey prior to 1993/1994 did not include any data from the territories. Data from the Northwest Territories were first added to the HMDB in 1993/1994.

As of data year 1994/1995:

At CIHI, data accuracy is measured in terms of unit non-response, item non-response and measurement error.

Unit non-response

As of fiscal 1999/2000, acute care hospitalizations from all provinces and territories are included in the HMDB. The Yukon Territory was added in fiscal 1994/1995 and Nunavut in 1999/2000. In fiscal 2002/2003, however, Nunavut did not submit data to the DAD and consequently Nunavut is not included in the fiscal 2002/2003 HMDB.

All known acute care facilities in Canada are routinely included in the HMDB. Since 1994/1995, however, inclusion of records from non-acute hospitals has been decreasing.

Item non-response

Item non-response is the degree to which mandatory data elements have not been completed on a hospitalization record. Where DAD is the source of the hospitalization records, uncorrected missing values and invalid data are assigned a standard default value (for example, 'Z'). Typically, only 1.2% of all records contain defaulted values. For hospitalizations from other sources (for example, Quebec) the item non-response rate. For example, Quebec submits the patient's age but not a birth date and submits only the first three digits of the patient's postal code.

Measurement error

CIHI assesses the overall quality of the DAD by conducting re-abstraction studies from a sample of facilities and jurisdictions. This re-captured information is then compared to the originally submitted records to determine consistency in coding and adherence to rules and guidelines. These studies also monitored the implementation of the new diagnosis classification, the International Classification of Diseases, 10th edition, Canadian Adaptation (ICD-10-CA) and the procedure classification, Canadian Classification of Health Intervention (CCI). Documentation from CIHI regarding the re-abstraction studies carried out in fiscal 2001/2002 and fiscal 2002/2003 indicate that the new coding schemes have been well adopted, there are concerns regarding diagnosis typing standards (the level of significance assigned to each reported diagnosis) and adherence to specific coding standards. Overall, however, the studies have confirmed the strength and fitness for use of the database.

CIHI reviews the results of re-abstraction studies and then implements solutions, such as workshops on special topics for coders in hospitals and improving coding guideline documentation.

Extensive documentation on the data quality of the Discharge Abstract Database, the Hospital Morbidity Database and the implementation of ICD-10-CA and CCI are available on the CIHI website.

Documentation