QUALITY CONTROL PROTOCOL

GENERAL RULES:

1. Keep ALL the OPD slides read by the lab in the previous month (see annex III: how to keep the slides).
2. Store the slides daily and separate the positive slides from the negative. The sample size calculation are done separately for negative and positive slides as the main measure of performance are:
   • Sensitivity, based on the total number of positive slides.
   • Specificity, based on the total number of negative slides.
3. If you have to send a sample of slides always take the sample randomly: DO NOT CHOOSE THE SLIDES.
4. Always send slides of the previous month to SMRU as soon as possible: Old slides of 2 or 3 months old should never be send as the number of slides unreadable is expected to increase with time.
5. Take good care when recording the QC form (date, code, result and lab tech name): If you record wrongly a code or a result then the recording mistake will be interpreted as a reading mistake.
6. The result must show the specie(s) found, the stage of development (T, S and/or G), and the parasitaemia with a percentage of RBC parasitised for all PFT +++ (3 +) and ++++ (4 +). Note also the presence of pigment in WBC when found.
7. Avoid as much as possible to send slides through post. If you have to, prepare a good secure packing.

THE NUMBER OF POSITIVE SLIDES AND NEGATIVE SLIDES YOU HAVE TO SEND IS RELATED TO DIFFERENT FACTORS:

• The precision you want for your sensitivity and specificity value.
  The smaller the precision the bigger the sample taken will be. We should aim for a precision of at most p ± 0.03 with a 95 % confidence interval (level of significance α = 0.05).
  Example:if your sensitivity is p = 0.94 (94 %) with a precision p ± 0.03, it’s means that the real sensitivity value is comprised between 0.91 and 0.97 with a 95 % confidence interval.
• The expected values for the sensitivity and specificity: the expectation taken into account will be your last QC results. Use the following table to calculate the size of the sample needed.
• The number of slides invalid (unreadable), which cannot be cross-checked because of deterioration through time. This number varies especially with the season. All sample sizes in the table have been increased by 20 % to take this into account. If more than 20 % of slides are expected to be invalid the sample size should be increased accordingly.

FREQUENCY OF THE CONTROL

• It is related to the quality of the laboratory (previous QC) and the turnover of the staff: a good laboratory (sensitivity & specificity ≥ 95 %) with stable staff and continuous supervision of the smears and stains quality, needs a QC once or twice a year.

ANALYSIS
The following indicators should be assessed.

• Smears quality: thick smears, thin smear and stain.
• The global percentage of errors.
• Sensitivity, Specificity, Predictive positive value (PPV), Predictive negative value (PNV). All these values should be given with a precision (see annex I).
• Kappa value. It gives a range of agreement (from poor to very good) in the species differentiation (positive slides) between the lab and the control. This value should be given with a precision (see annex II).
• The detection accuracy of the different development stages (T, S & G). This is important especially for Plasmodium falciparum.
• The parasitaemia accuracy.

Important:

For the analysis of your QC we need the total number of slides positive and the total number of slides negative read by your lab during the dates matching the sample of slides sent for QC. Please send these 2 numbers altogether with the slides.

---

TABLE: Sample of POSITIVE & NEGATIVE slides for QC

N _-ve = total number of Negative slides read during the month
N _+ve = total number of Positive slides read during the month

p₁= Specificity or Sensitivity expected p₁ ≥ 93 %
p₂= Specificity or Sensitivity expected 87 % ≤ p₂ < 93 %.
p₃= Specificity or Sensitivity expected 82 % ≤ p₃ < 87 %.

N -ve or N +ve	< 110	150	200	250	300	350	400
p1	All slides	110	130	145	155	155	170
p2	All slides	130	170	190	220	230	250
p3	All slides	All slides	180	220	240	265	290

N -ve or N +ve	500	800	1000	1500	2000	2500	3000
p1	170	205	205	215	230	230	230
p2	275	325	350	380	410	420	430
p3	320	400	430	480	530	550	580

Count N _-ve and N _+ve separately for the past month.

Take into account your sensitivity expected (last QC) for the number of positive slides to be taken. Take into account your specificity expected (last QC) for the number of negative slides to be taken.

---

Example 1:

• Last month your Total number of positive slides was 220. Your last QC showed a sensitivity of 92 %.

How many positive slides do you need to send for QC?

• Look into the second row: p₂ (Sensitivity expected 87 % ≤ p₂ < 93 %). N _+ve (220) is roughly between 200 and 250, so the number of positive slides to be sent is between 170 and 190: 180.
• Important remark: this number is an estimation for a significant sample size. It does not need to be respected exactly. If you are 5 less or 5 more, that does not matter.

How to select them randomly?

• If you have 220 positive slides and you want to send 180, you need to take off 40 slides randomly.
  40 slides out of 220 (40 / 220 = 0.182) is close to 1 out of 6 (1 / 6 = 0.167). Therefore, without choosing the slides, you will take off 1 slide every 6 slides: 183 will remain to be sent for QC.

Example 2:

• Last month your Total number of negative slides was 613. Your last QC showed a specificity of 98 %.

How many slides do you need to send for QC?

• Look into the first row: p₁ (Specificity expected p1 ≥ 93 %). N _-ve (613) is between 500 and 800, so the number of negative slides to be sent is between 170 and 205: about 190.

How to select them randomly?

• If you have 613 negative slides and you want to send around 200, you need to keep 1 slide out of 3 randomly (200 / 600 = 2 / 6 = 1 / 3).
  Therefore, without choosing the slides, you will keep 1 slide every 3 slides for QC.

---

ANNEX I: SENSITIVITY, SPECIFICITY, P.P.V, P.N.V CALCULATION & PRECISION.
Example:

• Total number of slide positive read by the lab in that month: 102
  
• Total number of slide negative read by the lab in that month: 383

Positive / Negative cross-tabulation.

		SMRU control
		+	Neg	Total
L	+	99	3	102
A	Neg	6	174	180
B	Total	105	177	282

Sensitivity: 94,3 % (99 / 105) P.P.V: 97,1 % (99 / 102)
Specificity: 98,3 % (174 / 177) P.N.V: 96,7 % (174 / 180)

Precision: p ± 1.96 x se

p is the proportion: sensitivity, specificity, P.P.V or P.N.V

1. Sensitivity: p = 99 / 105 = 0.943
   
   The sensitivity is related to the number of slides positive. All the positive slides were crosschecked (no sample taken randomly) therefore the sensitivity value is the reel one and there is no precision value.
   
   Sensitivity = 94.3 %
   
   
2. Specificity: p = 174 / 177 = 0.983
   
   Npop = N -ve = Total number of negative slides read by the lab in the month =383.
   
   N = Denominator for the specificity =177.
   
   se = 0.007
   Specificity = 0.983 ± 0.0139
   Specificity = 98.3 % precision 95 % CI (96.9 %, 99.7 %)
   
   
3. P.P.V: p = 99 / 102 = 0.971
   
   The P.P.V is related to the number of slides positive. All the positive slides were crosschecked (no sample taken randomly) therefore the P.P.V is the reel one and there is no precision value.
   
   P.P.V = 0.971 = 97.1 %
   
   
4. P.N.V: p = 174 / 180 = 0.967
   
   Npop = N -ve = Total number of negative slides read by the lab in the month =383.
   N = Denominator of the P.N.V = 180.
   
   se = 0.0097
   P.N.V = 0.967 ± 0.019
   P.N.V= 96.7 % precision 95 % CI (94.8 %, 98.6 %)
   

Particular case:
When sensitivity, specificity, P.P.V or P.N.V = 100 %, Then p = 1 and the above formula is inapplicable.

Then the precision is:

α1N, 1

• α is the level of significance, α = 0.05
• n is the sample size

Example: Specificity: p = 143 / 143 = 1

• α^1N = 0.05 ^{1/ 143} = 0.979
  

Specificity = 100 % precision (97.9 %, 100 %)

---

ANNEX II: KAPPA CALCULATION & CONFIDENCE INTERVAL.
Kappa (K) is a statistical value used to quantify the agreement between results.
Kappa ratings are as follow:

• 0 ≤ K ≤ 0.2 : poor agreement
• 0.2 < K ≤ 0.4 : fair agreement
• 0.4 < K ≤ 0.6 : moderate agreement
• 0.6 < K ≤ 0.8 : good agreement
• 0.8 < K ≤ 1 : very good agreement

Example:

Species (positive / positive) cross-tabulation.

		SMRU Control
		Pf	Pf+Pv	Pv	Pm	Total
L A B	Pf	74	2	0	0	76
	Pf+Pv	0	5	0	0	5
	Pv	2	1	72	0	75
	Pm	0	0	1	4	5
	Total	76	8	73	4	161

p0 = proportion observed agreement.
pe = proportion expected agreement.

Precision: K ± 1.96 x se

K = 0.934 ± 1.96 x se = 0.934 ± 0.052

Kappa = 0.934 Precision (0.882, 0.986) Very good agreement.

---

ANNEX III: HOW TO KEEP AND STORE THE SLIDES IN GOOD CONDITION.
Store and keep slides in good condition is mainly a fight against humidity (fungus growth) and dust.
The fight against humidity is a difficult challenge in the peripheral laboratories on this border, especially during the rainy season.
The following points might help:

• To keep a slide in good condition, it has to be in good condition in the first place: smearing, drying, fixing and staining techniques must be correctly performed.
• Before starting to read the slides place beside the microscope pre-cut bands of newspaper (about 8-cm width).
• Put the MINIMUM quantity of immersion oil necessary for the reading.
• When the reading is finished, remove the slide from the stage and put it FACE DOWN (protection from dust) on the news paper band. Start to separate positive slides from negative slides at this point: one newspaper band for the positive and another one for the negative.
  The newspaper will absorb the oil from the slides.
• When the work of the day is finished, take a NEW band of newspaper and roll the slides into it (separate the positive from the negative).
• Then, pack the slides in a plastic bag with zip and record the date and "positive" or "negative" slides on the bag.

Note: this method is one among others and does not pretend to be the best. However it is one of the cheapest.
It is also possible to store slides in hermetic boxes with silicate gel or to send the slides regularly to a suitable place with air-conditioned.

---

QC form (2 MS Word 2000 documents) in Zip