Concept: Follow-Up Issues

Introduction

This concept discusses the various threats to follow-up and how to resolve them.

Follow-up is the process of determining whether a member of a known cohort/population is in fact resident and eligible for Manitoba health care coverage for the entire duration of a study or until death. Follow-up is occasionally defined according to the nature of the proposed analyses. As a rule, individuals born during the period of study and those who die before the study's latest service date are not considered to be 'lost to follow-up' since all services are available to the study.

Differential follow-up may affect the generalizability of the results. This is a particular issue in studies of the long-term outcomes of "infant health". Differential follow-up is also a threat to the generalizability of sibling and twin research. Comparisons using whole sample and sibling data are very useful for exploring generalizability (i.e.: the degree of follow-up for these "special" samples may differ from that of the whole population) and bias. For example, whole sample and sibling analyses indicate no important effects of "infant health" on migration out of Manitoba (Oreopoulos et al., 2006 ). Once family fixed effects (using sibling data) are controlled for, estimates of the impact of "infant health" on later outcomes among Manitoba residents do not appear to be biased due to children with particular characteristics leaving the province.

Threats to Follow-Up

Newborns often introduce confusion due to the presence of the birth and neo-natal services that may inflate admissions and days of stay relative to a cohort that excludes first year of life. Delays in registration of newborns also make it difficult to resolve threats to follow-up.

One possible outcome, mortality, may or may not represent a threat to follow-up. Occasionally, persons known to be deceased may be excluded from a cohort-generally when the study proposes to initiate contact (such as a survey in collaboration with Manitoba Health) or assess the few years of a specific program, etc.



1. Migration / Interprovincial Migration



• out of Manitoba - left province before the study's latest service date
  
  
• to Manitoba - arrived in province after the study's earliest service date


Note: in either case, the health history of such individuals is unknown before arrival or after departure.



Exception: For the first three months of residence in Manitoba, individuals retain coverage in their province of origin, and for the first three months following departure from Manitoba for a reciprocating province, they retain coverage in Manitoba. Most of the health history for such individuals is captured in the Manitoba files regardless of residence or province of coverage.

In both circumstances most services will be reported in Manitoba's medical services/physician claims, reciprocal services and/or hospital discharge abstracts data.


2. Non-continuous eligibility - any periods of residence during which health services are the responsibility of an authority other than Manitoba Health - i.e. military, prison, RCMP, ward of the province, and some areas related to status First Nations bands. NOTE: Prior to April 1, 2013, members of the RCMP living in Manitoba were insured federally and not included in the Manitoba Health Insurance Registry data. After this date members of the RCMP living in Manitoba are covered by Manitoba Health and are included in the Insurance Registry data.

3. Multiple registrations - failure to cancel previous registration when departing Manitoba and subsequent re-entry from a non-reciprocating jurisdiction



• for dependents (before they reach age 18 years) or spouses, this applies as long as the original family head remains registered or until the change is reported to Manitoba Health.
  
  
• for family heads, this applies within two years of most recent service for any family member


The open coverage period may continue until the family corrects their health card or whenever MB Health tries to issue the person a revised personal registration card.



• above situation compounded by intervening marital and surname changes.
  
  
• notification of birth by vital statistics versus the parent(s) results in coding differences. May also occur when there are birth twins, or when early death reported before the birth; the death report could be ignored due to lack of registration. Either situation could result in a false positive (i.e. a record is left open when it should be closed).
  
  
• Child & Family Services interventions to avoid leaving an electronic record of adoption or residence, the previous registration is cancelled and a new one is issued using a different registration number (REGNO) / (REGNO_CODE) and PHIN.


Note: An individual for whom multiple registrations have been detected will also have two or more PHINs linked to the active PHIN.

4. Identification problems due to coding error



• at source - coder/provider might attribute services to the wrong patient.
  
  - coder/provider might confuse multiple siblings
  
  - coder/provider might report ineligible persons using valid registration number
  
  - coder/provider might report eligible persons using an expired registration number.
  
  
• at MB Health - from 1984-1994 keying of PHIN was done semi-automatically based on reasonable matches of names and characteristics. Siblings and consecutive spouses could be assigned incorrect (but legitimate) PHINs.

  Note: Since 1994, PHIN is increasingly captured from claims but the validity checks are still prepared.

• in Drug Program Information Network (DPIN) data - coder error in entering PHIN at point of sale (e.g. if there are multiple siblings, the medication could be attributed to the wrong one).
  
  
• anywhere - coding of different names, spelling, birth dates etc. can isolate some claims from others for the same individual. The registry can help determine the likelihood that each observed value (set of identifies) represents a known registrant; this process will also provide a single unique identifier for use in future analyses.


5. Loss to follow-up - at some point during the study period a cohort member becomes ineligible for provincial health care coverage because of:



• migration out of Manitoba with a known or unknown departure date
  
  
• migration to Manitoba, birth or otherwise becoming eligible after the earliest date of the study period
  
  
• registration in error - have known cancellation dates
  
  
• cancellation for institutional reasons (e.g. separate federal funding for members of the military, the RCMP, etc.) - have known cancellation dates.

Follow-up cannot be applied if:

• The cohort/population being studied includes services provided in Manitoba for non-residents; mortality and continuity issues cannot be resolved for these individuals.
  
  
• Comprehensive population registries are not available for the period of interest.
  
  
• The study is using sources that cannot be linked to Manitoba registrations due to missing or unreliable standard identification fields (such as PHIN, registration number, or postal code, along with sex, birth date, and name initials).

Follow-up issues should be resolved when:

• A study proposes to define a cohort from any source and study utilization over any period of time. It will require: identity validation, mortality, migration, ineligibility checks
  
  
• A study is defined based on physician practices that include measures related to variation in workload or continuity of care for individual patients.
  
  
• A study is doing cross-sectional analyses for any period longer than one day and requires measures related to the number of individuals involved (other than patient/days or 'N' services). The fiscal data may include individuals appearing at two or more identifiers. These should be resolved prior to analyses.

Geographic Mobility

This refers to the movement of a Manitoba resident within the province that results in a change of postal code or municipal code. Though it does not affect continuity of eligibility (follow-up) it may introduce changes in income quintile over time due to changes in postal code.

Analyses using regional cells may show individuals appearing in more than one cell over time. This is acceptable in many instances (e.g. physician workload or costs, institutional workload or costs, regional incidence of disease). A problem that may develop is that the number of individuals summed over all regions can exceed population denominators.

It is also possible to assign service data to a single region per individual using the earliest or most recent residence items observed OR as of a given date/event. This option raises other issues that must be addressed by the research team.