Introduction

In addition to prevalence and incidence measures, for many pharmaceutical-based projects, we need to know not only when a drug is started but also how long a person is exposed to the drug. Usually, we try to create a time window of probable use or exposure, because this window has many pharmacoepidemiologic applications. Consider the following (very complex) question:

Is there a dose/response and/or duration effect of highly-anticholinergic antidepressants vs. less anticholinergic antidepressants on incidence of hip fracture in community-dwelling elderly?

It is an example of a pharmacoepidemiologic question where one would need to codify and examine:


• dose of antidepressant (DDD and/ or PDD)
  
• duration of use (medication total exposure calculation)
  
• different kinds of antidepressants (finding the 'right' drug), and
  
• hip fracture diagnoses (ICD-9-CM)

Measuring duration of use or a reasonable 'time window of probable pharmaceutical use' requires some basic calculations of medication intervals:


1. Days in a medication-taking interval [MED_INT] ; similar to defining the length of an 'episode of care'
2. Proportion of days on medication [MED_TOTAL] ; during a medication-taking interval (MED_INT)
3. Sum of days without medication [TOTDWOUT] ; calculated ONE medication-taking interval at a time
4. Proportion of days not on medication [MED_OUT] ; during a medication-taking interval (MED_INT)
5. Sum of days with too much medication [TOTDOVER] ; calculated ONE medication-taking interval at a time
6. Proportion of days on medication [INT_COMP] ; during ONE medication-taking interval

EXAMPLE: IBUPROFEN 400mg

Consider the following record of consecutive dispensations of ibuprofen 400 mg (NSAID) to a person in the Manitoba Health Registry in interpreting the duration of use calculations and measures:



Time in Days	Date1	Date2	Diff in Days	DS	QTY	INT-COMP	Days W/O	Days Over
0	1/6/96	2/22/96	47	30	100	0.64	17	0
47	2/22/96	3/20/96	27	30	120	1.11	0	3
74	3/20/96	4/17/96	28	30	120	1.07	0	2
102	4/17/96	5/23/96	36	30	120	0.83	6	0
138	5/23/96	6/25/96	33	60	120	1.82	0	27
171	6/25/96	7/18/96	23	30	100	1.30	0	7
194	7/18/96	-	-	25	100	-	-	-
		TOTAL	194	210			23	39
		MED-INT	219 (194+25)
		MED-TOTAL	0.96 (210/219)
		MED-OUT	0.11 (23/219)

1. Medication Interval [MED_INT]

Calculation: The number of days from the first dispensation (i.e., index date) to the last day of dispensation (i.e., study end date, date of death, date of event or emigration from the province, thereby ending provincial insurance registration).

This is similar to defining the length of an 'episode of care'; often used to define "chronicity", persistence and concurrency.

Example:

First prescription (Rx) dispensation: 06 Jan 1996
Last prescription (Rx) dispensation: 18 Jul 1996
Last Rx day supply 25 days
Medication interval (MED_INT) : 194 + 25 = 219 days

See SAS code.

2. Medication Total [MED_TOTAL]

Defined as the proportion of days on medication during a medication-taking interval (MED_INT).

Calculation: The sum of the stated number of days of medication supplied (DAYSUPP) over the entire medication interval (06 Jan 1996 to 18 July 1996). Medication total (MED_TOTAL) should be about 1.0 which would approximate and assume 'full' compliance/adherence to the directions for use on the prescription.

Example:

Sum of days of medication supplied (DAYSUPP): 210
Medication interval (MED_INT): 219
(Assumed) proportion of days on medication (MED_TOTAL): 210/219 = 0.96 ((Sum DAYSUPP)/MED_INT)

See SAS code.

3. Sum of Days Without Medication [TOTDWOUT]

Calculation: Days without medication (W/O) during ONE medication-taking interval (difference between the actual days in the dispensing interval and the days supply, if negative, the value is a 'days without' value); the sum (S) of days without (TOTDWOUT) over the course of a medication interval is used in the calculation of proportion of days without medication (MED_OUT).

Example:

Days in Rx interval: 06 Jan 1996 to 22 Feb 1996: 47
Days supplied for this prescription (Rx): 30
Difference: 17
Sum of days without medication (including the 6 days from 06 Jan 1996 to 18 Jul 1996) : 23

This is an interim calculation needed to determine how many days a person was likely without sufficient medication supply (MED_OUT) to have a therapeutic effect. It is done on a dispensation to dispensation basis and the differences between days in the interval and days supply over a person's history of dispensations are summed. It is 'twinned' with the calculation of days over on medication (TOTDOVER).

(In other words, the total number of days with medication is 194 and the net of days without and days over is 16 or 210 days supplied in total. However, if one adds the last days supplied amount to what we know is the number of days of possession (n=194 days) then one is left with an approximation of medication interval (MED_INT) (n=219 days).

See SAS code.

4. Medication Out [MEDOUT]

Defined as the proportion of days not on medication during a medication-taking interval

Calculation: The sum of days without medication over the entire (estimated) medication interval (06 Jan 1996 to 18 July 1996). Medication out (MED_OUT) should be about 0.0 which would approximate and assume 'full' compliance/adherence to the directions for use on the prescription.

Example:

Sum of days of medication without (TOTDWOUT): 23
Medication interval (MED_INT): 219
(Assumed) proportion of days not on medication (MED_OUT) 23/219 = 0.11

(TOTDWOUT / MED_INT)

See SAS code.

5. Sum of Days Over With Medication [TOTDOVER]

Calculation: Days over with medication during ONE medication-taking interval (difference between the actual days in the dispensing interval and the days supply, if positive, then value is a "days over" value); the sum (S) of days over (TOTDOVER), over the course of a medication interval, is used in the calculation of proportion of days without medication (MED_OUT).

Example:

Days in Rx interval: 23 May 1996 to 25 Jun 1996: 33
Days supplied for this prescription (Rx): 60
Difference: 27
Sum of days over with medication (including 12 days from 06 Jan 1996 to 18 Jul 1996): 39

This is a companion calculation to the total days without calculation (TOTDWOUT). It is done on a dispensation to dispensation basis and the differences in the interval and days supply over a person's history of dispensations are summed. It is 'twinned' with the calculation of days without medication (TOTDWOUT).

In other words, the total number of days with medication is 194 and the net of days without and days over is 16 or 210 days supplied in total. However, if one adds the last days supplied amount to what we know is the number of days of possession (n=194 days) then one is left with an approximation of medication interval (MED_INT) (n=219 days).

See SAS code.

6. Interval Compliance [INT_COMP]

Defined as the proportion of days on medication during ONE medication-taking interval.

Calculation: The days supplied (for one dispensation) divided by the number of the days in dispensation interval. Interval compliance (INT_COMP) should be about 1.0 which would approximate 'full' compliance/adherence to the directions for use on a prescription.

This calculation gives an indication of how the person actually took the medication and is often used to determine if a prescription's dose being dispensed, over time, falls below that needed for therapeutic effect.

Example:

Days in Rx interval: 22 Feb 1996 to 20 Mar 1996: 27
Days Supplied for Rx dispensed 22 Feb 1996: 30
(Assumed) proportion of days on medication during one medication interval (INT_COMP) 30/27 = 1.11 (DAYSUPP/MED_INT)

See SAS code.

Applying Duration of Use Calculations

Since the Drug Program Information Network (DPIN) data does not include 'reason' the physician prescribed the drug, it is necessary to use other means to determine 'indication for use'. In one study, for example, examining NSAIDS for osteoarthritis and rheumatoid arthritis (MAAUI) (Speedie et al, 1992; Leufkens et al, 1990) , indication for use was approximated with the following calculations of 'duration of use', characterizing use of a drug as episodic, chronic, or intermittent.

(1) Episodic: Use according to the following algorithm approximated those taking NSAIDs for short-term musculoskeletal problems like a torn hamstring or sprained ankle:

Calculation: In one year (01 August-31 July), one only prescription or two to three prescriptions with a total duration of use of less than 100 days; total days without medication of = 61 days (TOTDWOUT) and MED_OUT (i.e., sum of days without medications / sum of days in all refill intervals) = 0.61 (about 0.0 if taking drug more than episodically)

(2) Intermittent: Use according to the following algorithm approximated those taking NSAIDs intermittently for neurovascular (e.g., menstrual headache) or gynecological problems (menstrual cramping).

Calculation: In one year (01 August-31 July), total days without medication of between 31 and 60 days (TOTDWOUT), MED_TOTAL (i.e., sum of days supply dispensed / sum of days in all refill intervals) between 0.4 and 0.79 and MED_OUT (i.e., sum of days without medications / sum of days in all refill intervals) between 0.21 and 0.61.

(3) Chronic: Use according to the following algorithm approximated those taking NSAIDs episodically for osteoarthritis or rheumatoid arthritis.

Calculation : In one year (01 August-31 July), total days without medication of = 30 days (TOTDWOUT), MED_TOTAL (i.e., sum of days supply dispensed / sum of days in all refill intervals) = 0.8 and MED_OUT (i.e., sum of days without medications / sum of days in all refill intervals) = 0.2.

The definition of chronic use depends on the type of drug being studied. In general, a person must have been dispensed two (2) or more prescriptions and be on medication for a percentage (x%) of time. The x is dependent on the drug therapy. For example, NSAID users had to have been taking a NSAID 80% of the time to be defined as a 'chronic user'.