Residence Determined by Reported Residence and
Geocoded Residence in the Michigan Death & Birth Certifcate Files

Using a process called geocoding, the Division for Vital Records & Health Statistics (DVRHS) has assigned geographic coordinates to most death certificates reported between 2000 and 2014. While the primary purpose of geocoding mortality data is to map patterns of death and to conduct spatial epidemiology, such records can also be used to validate location-related data recorded on the certificate. In particular, validating reported minor civil division (MCD) of residence has become important due to increasing interest in small area statistics. In other health-related datasets, the division has found that the MCD residence is miscoded in certain areas about 10-20% of the time. Often, the person reporting the residence confuses the township with an adjacent city or village. Such errors then alter the calculation of local area statistics; small errors in coding lead to large errors in rates calculated for MCDs with populations below 70,000.

Therefore, the DVRHS studied the accuracy of the minor civil division (MCD) of residence reported on the death certificate. As part of the validation, the residence was compared to the geocoded minor civil division (MCD) using a "comparability ratio." Note that, the comparability ratio is defined as the three year average of the number of deaths geocoded to a particular MCD, divided by the number of deaths reported for the MCD. A comparability ratio of 1 indicates that the geocoded and reported codes are statistically equivalent. The greater the absolute difference of the comparability ratio and one, the less accurate the reported codes are. For example, a ratio of .97 and 1.03 indicates that the reported and the geocoded-derived code almost always agree. A ratio or .8 or 1.2 indicates that the reported codes are inaccurate.

Table B shows comparability ratios for MCDs with populations over 10,000. The ratios for 60% of the MCDs are close to 1 (1 ±.06), indicating that the decedent's residency was reported accurately. (A smaller version, without the death counts, is shown to the right, Table A.) This accurate group of MCDs includes all the cities and townships in Michigan with population over 70,000. An additional 26% of the reported MCDs were moderately correct (the ratio was within the band (.82 , .94) or (1.06, 1.18)), but the reporting error contributed to the overall instability of the rates. Age-specific and age-adjusted rates could not be accurately calculated for these MCDs from the reported residence. However, the counts of deaths from this "moderate" group were still useful for calculating sample statistics such as percents. The remaining 15% of MCDs were inaccurate and propagated errors throughout any statistical calculation.

Normally, DVRHS concludes geocoding about six months after it publishes mortality statistics on the web. Prior to this release, the division provides MCD statistics primarily through the Community Health Information web site. The division has modified this website to provide statistics by geocoded MCD for years when the data is available. However, there are periods when geocoded data is unavailable. In these cases, counts and rates for MCDs with inaccurate reported residence codes will be censored; the website will report the previous year's data instead. A list of these MCDs can be identified by submitting ">±.06" to the Table A form (right).

If the residence information is accurate, the mortality statistics are provided using the reported residence from the death certificate and then, the geocoded residence as the data becomes available. Essentially, for all cities in Michigan, with population over 70,000, mortality statistics for the most current release of data will always be provided; for other MCDs with inaccurate data, there will be about an additional six month delay.

Loading. Please wait...