Technical Notes

Interpreting the Measures

Plan-specific Rates

The majority of rates reported as part of the 2010 Report on Managed Care Performance: A Report on Quality, Access to Care and Consumer Satisfaction are displayed as rates per 100 (percentages). To calculate a plan's rate for a measure, the numerator is divided by the denominator and then multiplied by 100.

Plan-specific data are excluded from the tables as a result of any of the following methodological limitations:

  • The denominator is less than 30, resulting in an unreliable rate. Please note that even though the plan's sample is too small to report individually, the plan's rate is included in the statewide average.
  • No enrollee could meet the eligibility requirements (such as continuous enrollment).

Prenatal Care

As in previous publications of QARR, several measures are calculated using both member-level information on live births that are submitted by the plans and the department's Vital Statistics (VS) birth file. The plans' records are matched to the VS data to find the most accurate numbers to perform the calculation. However, if a record is missing "Trimester Prenatal Care Began" on the VS birth file, that record is excluded from the calculation. The reporting of this information is the responsibility of the hospital of delivery. As a result of the exclusion, plans' rates are less likely to be affected by the hospital's failure to report complete birth data.

Risk-Adjustment Factors

Health events, such as low birthweight (LBW) births do not occur randomly across all plans. In addition, certain risk factors, such as maternal age or education, may be disproportionate across plans and beyond the plans' control. Risk adjustment is used because it removes or reduces the effects of confounding factors that may influence a plan's rate. These data reflect the removal of multiple births and include only women who were continuously enrolled in a plan for ten months, allowing for a one-month break in service. Therefore, risk-adjusted rates account for patient factors that strongly influence the outcome, thereby allowing for a fairer comparison among the plans.

Low Birthweight Methodology

To compute the risk-adjusted LBW rates, a logistic regression model was developed. The model predicted a binary response for LBW, i.e., all births were designated as either LBW or "not LBW" (<2500 grams).

The independent variables used in the methodology included:

  • maternal age (less than 18, 18-19, 20-29, 30 and over)
  • education (less than high school, high school, any college)
  • alcohol use (yes, no)
  • drug use (yes, no)
  • tobacco (yes, no)
  • level of prenatal care as defined by a modified Kessner index (intense, adequate, intermediate, inadequate, no care, unknown)
  • race/ethnicity (white, black, Hispanic, other)
  • parity (none, 1-2, 3-4, 5 or more previous live births)
  • resident of New York City (yes, no)
  • maternal medical risk factors (yes, no)
  • previous low birthweight (yes, no)
  • previous pre-term delivery (yes, no)
  • nationality (born in US/Puerto Rico or rest of world)
  • marital status (yes,no)
  • Medicaid aid category (ADC, Safety Net, MA, SSI, FHP)
  • hospitalized during pregnancy (yes,no)
  • prelabor referral for high risk pregnancy (yes,no)
  • vaginal bleeding (yes,no)
  • How did you feel about the timing of pregnancy (sooner, later, now, never)?

The expected LBW rate is the rate a plan would have if the plan's patient mix were identical to the patient mix of the state. The plan-specific, risk-adjusted rate is the ratio of observed to expected LBW rates multiplied by the overall statewide LBW rate.

Limitations of the Risk-Adjusted Data

The Risk-Adjusted methodology allows for more accurate comparisons among plans. Nevertheless, it has some limitations. The information on the Vital Statistics Birth File is reported by hospitals and is not validated or audited for accuracy. Therefore, inaccuracies in birth certificate data may influence the risk-adjusted rates. Also, if important risk factors are not included in the model as independent variables, the model can potentially overestimate or underestimate a plan's risk-adjusted rate. Although the limitations presented here are an important consideration in interpreting the risk-adjusted data, comparisons between plans are much more accurate when using this data, than if non-adjusted data were used.

Need More Information

If you have any questions or comments about the Technical Notes section please contact the Bureau of Quality Measurement & Improvement at (518) 486-9012 or e-mail