Table 4. Estimated multivariate polynomial distributed lag (PDL) model for monthly %MRSA (R²=0.902)^a.

Explaining variable	Lag (mo.)	Direct effectb			Indirect effectc	Sum of both effectsd
		Coeff	T-stat	p	Coeff	Coeffe	T-stat	p
%MRSA	1	0.420	3.96	0.0003
Macrolide use
Each month	1	0.083				0.083	4.02	0.0003
	2	0.055			0.035	0.090	5.34	< 0.0001
	3	0.027			0.038	0.065	6.02	< 0.0001
	4				0.027	0.027	3.16	0.003
Overall	1–3	0.165	4.02	0.0003
	2–4				0.100
	1–4					0.265
Third-generation cephalosporin use
Each month	4	0.116				0.116	2.75	0.009
	5	0.087			0.049	0.136	3.27	0.002
	6	0.058			0.057	0.115	3.70	0.0007
	7	0.029			0.048	0.077	3.91	0.0004
	8				0.032	0.032	2.75	0.009
Overall	4–7	0.290	2.75	0.009
	5–8				0.186
	4–8					0.476
Fluoroquinolone use
Each month	4	0.170				0.170	3.43	0.002
	5	0.085			0.071	0.156	3.37	0.002
	6				0.066	0.066	2.31	0.03
Overall	4–5	0.255	3.43	0.002
	5–6				0.137
	4–6					0.392
Constant		–36.7	–4.42	0.0001

^aMRSA, methicillin-resistant Staphylococcus aureus.
^bPast %MRSA as well as past use of these three antimicrobial drug classes had direct effects on %MRSA. These direct effects diminished the longer the lag time.
^cBecause every increase in %MRSA by the value 1 was followed the next month by a significant increase in %MRSA by the value 0.420, use of the three antimicrobial drug classes also had indirect effects on the %MRSA. As 0.420 is <1, these indirect effects necessarily vanished over time. As an example, decreasing indirect effects are only presented for a few months. There were substantial indirect effects of macrolide use up to month 8 (final coefficient for sum of both effects = 0.284), of third-generation cephalosporin use up to month 12 (final coefficient for sum of both effects = 0.499), and of fluoroquinolone use up to month 11 (final coefficient for sum of both effects = 0.440).
^dEach month, the total effect of each class of antimicrobial on the %MRSA resulted from the sum of the direct and indirect effects.
^eThe estimated coefficients indicate the values by which the %MRSA would increase in response to an increase in 1 DDD per 1,000 patient-days for each of the three significant antimicrobial classes, when all other variables remain constant. Since the average figure for monthly patient-days at Aberdeen Royal Infirmary is 22,800, 10 DDD per 1,000 patient-days correspond to approximately 230 DDD per month or thirty 7- to 8-day antimicrobial courses. For example, an increase in macrolide use by 10 DDD per 1,000 patient-days on a certain month, or 30 more patients treated with a macrolide as compared with the previous month, would lead to a direct increase in %MRSA by 0.83, 1 month later, by 0.55, 2 months later and by 0.27, 3 months later. The total direct effect would therefore be evident after 3 months, amounting to an increase in %MRSA by the value 1.65. Additionally, %MRSA indirectly attributable to macrolide use would increase by the value 0.35 (i.e., 0.83 x 0.42) after 2 months and by 0.38 (i.e.. [0.83 x 0.42] + [0.55 x 0.42]) after 3 months. From the 4th month onwards, there would be no direct effect of macrolide use on the %MRSA, only ever-decreasing indirect effects that would practically disappear after 8 months (decreasing effects in months 5 to 8 not shown).